Short-Term Assessment of PM2.5 Physico-Chemical Characteristics at Three Different Sites in Indonesia and Their Potential Association with Respiratory Diseases

Study area and sampling procedure

Sampling was conducted in Bandung at Jalan Dr Djunjunan, Bandung City, in November 2021 for 24 hours. For the other location, sampling was conducted for 12 hours in Kecamatan Pedamaran Timur, Ogan Komering Ilir District, South Sumatra, in October 2021, and in Serpong City, Tangerang, in July 2022. Minivol Technical Air Sampler from AIR METRICS is used for PM2.5 sampling. MiniVol operates based on the impact force resulting from differences in air flow rate. When sampling particulate emissions, air is sucked in by the pump and enters through the size separator, where it is then trapped in the filter medium. Particulates carried in the sample air will be separated in the size separator based on size due to the impact force (collision) in the separator. The flow rate used is five litres per minute (LPM). The filter used for sampling was Tisch Scientific's PTFE membrane filter, 46.2 mm in diameter, with a support ring and a pore size of 2 µm. Before sampling, the filter was weighed first using a Sartorius microbalance. Weighing conditions are maintained by keeping humidity <50% and temperature ± 21 °C. Conditioning is carried out over a 24-hour period. After sampling, the filter is then weighed again. The filter is then stored in a special container before being analyzed. Then, the calculation is performed using the following formula (MiniVol Operation Manual):

Where:

PM2.5 = PM2.5 concentration (µg/m3)

PMs = filter weight after sampling (µg)

PMb = weight of filter before side (µg)

Vact = air volume when sampling (m3)

The effects of meteorological parameters were also studied using an Automatic Weather Station (AWS) instrument at three sampling sites. Ambient air temperature, total solar radiation, wind speed and direction, relative humidity (RH), and barometric pressure were measured. All calculations and graphics related to meteorology and associated with the PM2.5 transport process were carried out using WindRose and QGIS tools to investigate the geographical origins of atmospheric pollution.

Figure 1.

Sampling location

Morphological and elemental composition

The morphological and elemental composition of particulate matter samples was characterized using a scanning electron microscope equipped with an energy-dispersive X-ray spectrometer (EDX 20 XMax Oxford) operated at 20 kV in high-vacuum conditions (500 Pa) with a secondary electron detector. Squares (1×1 cm) of the filters for each studied site were cut and fixed to the sample holder using carbon tape gold, and then coated for 1 minute. The morphological and chemical parameters of the individual particles were obtained at different high magnifications, ranging from 5000 to 20000 times. EDS spectra were obtained from the single particle and a larger area in the sample to identify the significant constituents and trace elements. In this study, EDS weight percent calculations excluded oxygen, carbon, and fluorine from compositional normalization due to their dominance in the filter blank. The analysis, therefore, focused on trace element compositions that are potentially relevant to human health. The elemental percentages reported represent relative contributions within the trace (non-carbonaceous) fraction of PM2.5. Excluding elements associated with the filter substrate enables improved resolution and interpretation of a broader range of trace elements present in the collected particles, as also reported by Sevimoglu et al. (2025) and Prabhu et al. (2019).

Sample extraction

The aqueous extraction method determined water soluble ionic components (NH⁴⁺, Ca⁺², Mg⁺², Na⁺, K⁺, Cl^–, NO^3–, SO₄^–2) associated with sampled particulates. Each sample filter disc was cut into pieces and taken into a polypropylene tube (50 mL) containing 20 mL of deionised water. The tubes containing samples were sonicated for one hour, followed by a 1-hour settling period to allow the filter fragments to settle. Subsequently, sonication was continued for an additional 30 minutes. Finally, the extract was filtered through Acrodisc (0.45 μm) with a disposable syringe.

Water-soluble ion compositions

Extracted aqueous samples were subsequently analyzed with the Ion Chromatography instrument. Anions (Cl^–, NO^3–, and SO₄^–2) were analyzed using ICS 1500 DIONEX ion chromatography equipped with AERS 500 4 mm suppressor and conductivity detector. The stationary phase consisted of an IONPAC AS12 4x200 mm analytical column and an IONPAC AG12 4x50 mm guard column. The mobile phase was an eluent solution, a mixture of Na₂CO₃ 2.7x10^-3 M and NaHCO₃ 3x10^-4 M with a flow rate of 1.5 L/min. For cations (NH⁴⁺, Na⁺, K⁺, Mg²⁺, Ca²⁺) using ICS1600 DIONEX ion chromatography with a CERS 500 4 mm suppressor. The separation column used is an IONPAC CS12A, 4x250 mm, equipped with a CG12A 4x50 mm Guard Column. The eluent, as the mobile phase, was a 98% Methanesulfonic Acid (MSA) solution of 2x10^-2 M with a flow rate of 1 L/min.

Trajectory analysis of PM2.5 using the HySplit model

The Hybrid Single-Particle Lagrangian Integrated Trajectories (HYSPLIT) model, provided by NOAA, was employed to analyse air mass trajectories arriving at the sampling site using a backward trajectory approach. Meteorological data from the Global Forecast System (GFS) with a 0.25° spatial resolution were utilised, with air mass trajectories traced 3-7 days prior to the sampling day, including the two preceding days. The study employed a frequency cumulative analysis at 10 meters above ground level (AGL) to capture recurring air mass movements, incorporating vertical velocities to track the movement of air parcels based on atmospheric conditions. This method helps identify potential source regions contributing to air pollution at the sampling site.

Aerosol Optical Thickness from Himawari-8 satellite observation

In this study, due to the lack of field sampling data on PM2.5 over the Indonesian area, satellite data of Aerosol Optical Thickness (AOT) were utilized to evaluate the change in aerosol pollution levels at the provincial scale in Indonesia from 2018 to 2023 mainly in South Sumatra, West Java, and Banten, as the provinces where location of sampling site was conducted, and another seven provinces in Indonesia. Himawari is a geostationary satellite launched by the Japan Meteorological Agency (JMA) in 2014. The AOT product from the Himawari-8 satellite has a spatial resolution of 5 km and a temporal resolution ranging from hourly to monthly. In this study, the global AOT L2 Version 2.1 of the JMA product was used. The Himawari L2 monthly AOT data at integer UTC hours were downloaded from the Japan Aerospace Exploration Agency (JAXA) Earth Observation Research Center (www.eorc.jaxa.jp/ptree/terms.html) (open access).

Respiratory diseases prevalence data

The respiratory disease prevalence data used in this study consist of Acute Respiratory Tract Infection (ARTI) in all age groups and toddlers, pneumonia in all age groups and toddlers, asthma and asthma relapse, and also pulmonary tuberculosis (TB). The data were obtained at the provincial level for ten provinces in Indonesia, including the locations of the cities where sampling was carried out (South Sumatra, West Java, and Banten) from the Indonesian Ministry of Health through the five-year national health surveys: Riset Kesehatan Dasar (Riskesdas) 2018 and Survei Kesehatan Indonesia (SKI) 2023. (https://kemkes.go.id/id/survei-kesehatan-indonesia-ski-2023, https://layanandata.kemkes.go.id/katalog-data/riskesdas/ketersediaan-data/riskesdas-2018). This data collection was carried out through interviews and laboratory examinations. The diseases recorded based solely on the doctor's diagnosis (specialist and general practitioner) are Pulmonary TB (TBC) and asthma, as well as asthma relapse. For ISPA and pneumonia, the diagnosis is based on the diagnostic history of healthcare workers (specialists, general practitioners, midwives, and nurses). In addition to the healthcare workers' diagnostic history, ARTI and pneumonia were also assessed based on the symptoms they had experienced. The health prevalence data were used to provide supporting epidemiological context and to explore regional-level associations with measured PM2.5 characteristics, rather than to establish direct causal relationships. Pandemic-year data were not included due to data availability limitations, as publicly accessible respiratory disease data are only available for the five-year national surveys (2013, 2018, and 2023). Only health prevalence data from 2018 and 2023 were used in this study, as Himawari-8 aerosol observations were not available in 2013. This selection ensured temporal consistency between satellite-derived aerosol data and health indicators.

PM2.5 concentrations and meteorology

Table 1 shows the statistical summary of PM2.5 concentrations at three different sites: Bandung, OKI, and Serpong. Without considering meteorological conditions, it can describe the activity of PM2.5 sources in all locations. The high standard deviation value indicates that the PM2.5 concentration changed rapidly over a short period. The mean 24-hour PM2.5 concentrations in Bandung were 87.98 µg/m³, exceeding the National Quality Standard, as specified in Government Regulation No. 22 of 2021 (55 µg/m³ for a 24-hour measurement). The urban area of Bandung, with a very high traffic density, is a source of a high maximum 24-hour PM2.5 concentration, observed at 195.04 µg/m³ on 5 November 2021.

Table 1.

Statistical summary for PM2.5 concentration using Minivol Air Sampler in Bandung City (1-7 November 2021 (24 hours), n=7), OKI District (4-7 October 2021 (12 hours), n=7), Serpong (25-29 July 2022 (12 hours), n=10), and Bandung (27 June-1 July 2022 (12 hours), n=10)

For OKI and Serpong, the PM2.5 concentrations and standard deviations show lower values because the sampling duration was only 12 hours. OKI and Serpong exhibited lower maximum concentrations of PM2.5, which were observed at 90.40 µg/m³ and 62.03 µg/m³, respectively. Ihsan et al. (2023) reported that the average PM2.5 concentration measurement in Serpong in 2020 was 55.2 µg/m³ over a 24-hour sampling period, yielding a result similar to that of this study. Chandra et al. (2021) reported that the daily PM2.5 concentration in Bandung City, measured by low-cost sensors from August to September 2018, ranged from 40 to 110 µg/m³, which is consistent with the measurement result from this study.

The sources of PM2.5 at the three different locations also varied. In an urban area of Bandung, PM2.5 predominantly originates from vehicle emissions, soil and road dust, industrial emissions, biomass burning, and secondary aerosols (Lestari et al., 2009; Santoso et al., 2008). Serpong is a suburban area with significant sources of PM2.5, including diesel vehicles and industrial emissions, as well as additional contributions from oil and power plants, road dust, and biomass burning (Santoso et al., 2008). OKI is a rural area, and typical sources of PM2.5 emissions include biomass burning in peatland fires or slash-and-burn practices, land clearing for agriculture, and household activities (Wildayana et al., 2017).

Figure 2 illustrates meteorological variables, including wind speed, wind direction, and rain rate at each location during each sampling period, as well as their impact on PM2.5 concentration. Differences in PM2.5 concentrations among sites reflect the combined effects of emissions and meteorological conditions during each sampling time, rather than baseline pollution levels. The higher the wind speed, the easier it is to disperse and transport PM2.5 away from its source. Rainfall is the meteorological parameter that has the most significant influence on reducing PM2.5 concentrations (Li et al., 2017). The 24-hour PM2.5 concentration in Bandung tended to increase with lower wind speeds, indicating the influence of local sources (Fig. 2A and 2D). On the contrary, the 12-hour PM2.5 concentration in OKI and Serpong appears to increase with higher wind speeds, indicating the long transport of pollutants into the region (Fig. 2C and 2E).

Figure 2.

Concentrations of PM2.5, wind speed, wind direction, and rain rate at three sites: Bandung City (A, B) (November 2021, 24 hours sampling), OKI District (C, D) (October 2021, 12 hours sampling at day and night), and Serpong (E, F) (July 2022, 12 hours sampling at day and night).

In Bandung, the 24-hour average wind speed values range from 1 to 2 m/s. In OKI and Serpong, the 12-hour wind speed average values range from 0.12 to 0.76 m/s and 0.1 to 0.87 m/s, respectively, which are lower than in Bandung. The daily PM2.5 concentration in Bandung ranges from 41 to 195 µg/m³, with the highest value recorded on November 5, 2021. In November 2021, Bandung was still implementing Community Activity Restrictions (PPKM) during the COVID-19 Pandemic. However, the still high concentration of PM2.5 suggests that the possibility of community mobility or industrial activity remains high.

For OKI and Serpong, the sampling duration was 12 hours, day and night. The 12-hour PM2.5 concentration in OKI ranged from 14.3 to 90.4 µg/m³, with the highest concentration recorded on October 5, 2021 (at nighttime). In Serpong, the concentration ranged from 8.2 to 62.02 µg/m³, with the highest concentration observed on July 27, 2022 (during the daytime). The 12-hour PM2.5 concentration in OKI and Serpong exhibited a diurnal pattern, with higher concentrations observed during the daytime than at nighttime, primarily due to human activity as the dominant source of PM2.5, which decreased at nighttime. The wind speed value also showed a similar pattern. The finding is coherent with the diurnal pattern of PM2.5 in the dry season, which peaks during the day in Jakarta, Surabaya, and Pangkal Pinang (Cholianawati et al., 2024). Rain rate also became the dominant factor that can reduce the PM2.5 concentration (Arisanti et al., 2025; Cholianawati et al., 2024; Sinuraya et al., 2024; Ihsan et al., 2023; Istiana et al., 2023; Kusumaningtyas et al., 2021).

Figure 2B shows that the rain event in Bandung was intense, with a rain rate exceeding 5 mm, which directly decreased the PM2.5 concentration over the four-day sampling period. However, on days 5 and 6, an anomaly occurred, with a very high concentration of PM2.5 found on day 5. The high concentration on day 5 was attributed to biomass burning from local activity around the sampling location and fossil fuel burning from traffic activities. On day 6, the dominant wind direction came from the west (around 270 degrees), which was a residential area, and there was no direct influence from highway emissions.

In OKI and Serpong, rainfall events were very rare during the sampling day (Fig. 2D and 2F). The wind speed during the day was higher than at night, which was predominantly calm, so meteorological conditions at these two locations did not significantly influence PM2.5 concentrations dilution during the sampling period. Slower wind speeds, higher relative humidity, cooler temperatures, and lower planetary boundary layers are conducive to restricted PM2.5 transport, thereby reducing the rate of PM2.5 removal. The geography of the sampling site also affects the dispersion process of PM2.5 (Prabhu et al., 2019; Shankar et al., 2022; Ren et al., 2024).

Table 2 shows PM2.5 samples represented at every site for further chemical analysis. In this study, we used three to four representative samples. The PM2.5 concentrations were selected based on lower and higher values as outlined in the National Air Quality Standard, as stipulated in Indonesian Government Regulation No. 22 of 2021, as mentioned in the preceding section. The threshold is 55 μg/m³ for 24 hours of sampling, whereas the US EPA threshold is 35 μg/m³.

Table 2.

PM2.5 samples from each location analyzed in this study

Figure 3 shows the wind rose plot at every location during the sampling. Over a 24-hour sampling period, a westerly wind dominated the Bandung area. High wind speeds were observed in Bandung, ranging from 3.1 to 5.7 m/s, which became one of the factors affecting lower temperatures. In Serpong, the southwest and northeast wind directions dominate during the daytime. Meanwhile, the easterly wind was dominant at night. The daytime wind speed was around 2.1-3.1 m/s, higher than at nighttime, showing very calm wind conditions. The same characteristics of wind direction were observed in OKI during both daytime and nighttime, with higher northeast wind speeds during the daytime. The west, east, and northeast dominate the wind direction during daytime. The observed pattern indicates that pollutants tend to come from various directions (many sources) during the daytime and, conversely, become concentrated in specific areas due to calm winds and stable air conditions at night (Cholianawati et al., 2024; Cholianawati et al., 2022; Hamdi et al., 2023; Vecchi et al., 2007).

Figure 3.

The wind rose plot at three sites: Bandung City (1-7 November 2021, 24 hours of sampling), OKI District (4-7 October 2021, 12 hours a day and night of sampling), and Serpong (25-29 July 2022, 12 hours a day and night of sampling.

Morphological and elemental compositions of PM2.5

The SEM-EDX technique is advantageous in differentiating the sources of natural or anthropogenic particles. Particles from anthropogenic sources often have a spherical morphology due to industrial processes that occur at high temperatures. Regular shapes with a certain degree of facet symmetry usually come from natural sources, such as suspended dust or rock. The elemental composition of the particles, related to their morphology, was identified using the EDS method (Gonzales et al., 2016).

The SEM micrograph and the energy-dispersive X-ray Spectrometer (EDS) spectrum of the particles in PM2.5 were collected from three different sampling locations, as shown in Figures 4(a-d). The spherical, regular, and irregular shapes were observed in the micrograph of PM2.5 in Bandung City (Fig. 4a). In the OKI district, some soot particles were also found (Fig. 4b). Large particles with a diameter >1µm and some smaller particles were observed in Bandung and OKI. The elements Mg, Si, S, K, Ca, and Cu were detected in Bandung, whereas in OKI, the Ca element was not found. Si and S had the highest percentages observed in all locations.

Figure 4.

Micrograph SEM, EDX Spectrum, and elemental composition of PM2.5 samples from three different locations, Bandung (a), OKI (b), and Serpong (c).

Silicon-rich particles can be associated with natural sources due to the formation of silicon dioxide (SiO₂). Mg and Ca represent the crustal origins from the natural source and carbonate minerals, while K is often associated with biomass burning. Agro-industrial and vehicle activities are the dominant sources of S elements, as well as from long-distance transport (Frazin et al., 2020; Rodriguez et al., 2021; Popovicheva et al., 2016). The presence of Ca elements can be related to the formation of CaSO₄ as the product of the resuspension of the crust and the formation of the secondary aerosols emitted by biomass burning or fossil fuels from the chemical reaction between CaCO₃ and the SO₂ during the combustion process. Cu elements related to anthropogenic sources are primarily derived from industrial activities, such as exhaust emissions (Gonzales et al., 2016). Cu can also be emitted from the brakes and vehicle friction (Quijano et al., 2019).

The micrograph PM2.5 sample in Serpong (Fig. 4c) is dominated by soot particles with numerous spherical shapes. Soot particles or aggregates were commonly composed of ultrafine particles that aggregated together to form larger sizes or chains. This type of particle is primarily emitted from gasoline and diesel combustion, with elements dominated by Sulfur-rich particles and some other minor elements also present. The EDX spectrum reveals the elemental composition of the particles, which is dominated by Si, S, Mg, and K, indicating a contribution from natural sources, as also detected by Gao et al. (2018) and Pipal et al. (2011). Soot aggregates with S content were due to the conversion of gas to particles during the transport of the particles. These particles are known as the markers of anthropogenic emissions, primarily from industrial and vehicle sources. Soot particles are typically found on industrial sites, originating from combustion processes such as coal-burning factories, biomass-burning facilities, and power stations. Different types of fuels, incomplete fossil fuel combustion, industrial and residential combustion, biomass burning, and atmospheric processes affect the variation in the morphology of the soot particles (Gao et al., 2018; Pipal et al., 2014).

The elemental composition in all representative samples from three different locations is shown in Figure 5. Silica was the major element found in all samples from three locations. Mg and S were also detected in all samples and showed higher percentages than other elements, such as K, Ca, Fe, Cu, and Pb, at lower levels. The PM2.5 elemental composition in Bandung City reveals a significant concentration of Si in all three samples, ranging from 35% to 50%. S and Mg were also detected in all samples, with values of 10-15% and 15-20%, respectively. High concentrations of Pb were also found in one of the three samples, with a concentration of 12%. In contrast, Cu was detected in all samples, with a concentration range of 8-12%. K was found in only two of the three samples, with concentrations ranging from 2% to 4%, and Ca was detected in only one sample, with a concentration of approximately 4%.

Figure 5.

Elemental composition at three different locations. Bandung (BDG) 24-hour sampling (a), OKI District (OKI) 12-hour sampling (b), and Serpong (SRP) 12-hour sampling (c). Elemental composition is expressed as relative percentages of detected trace elements, excluding C, O, and F.

PM2.5 samples in OKI exhibit high concentrations of Si, Mg, and S in all four samples, with values ranging from 45% to 60%, 15% to 18%, and 15% to 20%, respectively. Pb was detected in one sample with a concentration of 10%, while Cu was detected in three samples with a concentration of 5-12%, and K was also found in all three samples with a concentration of 2-3%. The elemental composition of PM2.5 samples in Serpong detected Si at very high concentrations of 60-75%, which is higher than found in other locations. Mg was found in all four samples at 15-20%, which is similar to the values reported in different locations. S was also detected in PM2.5 samples in Serpong with a concentration of 10-15%, and K was only detected in one sample with a lower concentration of 5%. At the same time, other elements were not detected in the PM2.5 samples from Serpong.

Water-soluble ion composition

Figure 6 shows the water-soluble anion and cation composition in PM2.5 from three different locations. The total ion concentrations in Bandung, OKI, and Serpong ranged from 12.43 - 16.86 µg/m³, 11.74 - 25.05 µg/m³, and 9.93 - 18.75 µg/m³, respectively (Fig. 6a). The diurnal pattern of total ions was observed in Serpong, with a higher 12-hour total ion concentration during the day than at night. Higher daytime temperatures may facilitate photochemical activity and thereby promote secondary aerosol formation; however, the resulting aerosol levels depend on precursor availability and atmospheric conditions rather than temperature alone (Cheng et al., 2021; Alramzi et al., 2025). The percentage of total ion content in PM2.5 concentration was in the range of 22.33 - 31.66% for Bandung, 27.71-59.94% for OKI, and 41.49-78.66% for Serpong (Fig. 6b). Serpong has the highest nocturnal water-soluble ion content in PM2.5.

Figure 6.

Ion composition of PM2.5 (a) and % of the ion in PM2.5 samples (b) at three different sites: Bandung (BDG) 24-hour sampling, OKI District (OKI) 12-hour sampling, and Serpong (SRP) 12-hour sampling

Sulphate, nitrate, and ammonium (SNA) are the most abundant ions in almost every location. SNA were also the most dominant ions in some regions in China (Xiao et al., 2020; Wang et al., 2021; Guo et al., 2020). SNA are the secondary aerosols, and their formation in the tropical region is influenced by several key factors, including chemical reactions, meteorological conditions, and regional emissions (Gui et al., 2024; Kong et al., 2020; Pathak et al., 2009; Chen et al., 2022; Moravek et al., 2019; Choi et al., 2024). The highest concentrations of sulphate ion in Bandung, Serpong, and OKI reached values of 8.796 µg/m³ (sample code: BDG2-06), 7.662 µg/m³ (sample code: SRP11), and 11.308 µg/m³ (sample code: OKI004), respectively. Nitrate ions have a higher concentration in Serpong and OKI, whereas in Bandung, the concentration of nitrate is lower. The highest concentration of nitrate ions in Serpong was 3.308 µg/m³ (sample code: SRP15). In OKI, the highest concentration of nitrate ions was 3.099 µg/m³ (sample code: OKI004), while in Bandung, the nitrate ion concentration was lower than in other locations, reaching only 1.292 µg/m³ (sample code: BDG2-02). Higher concentrations of ammonium were found than those of nitrate in Bandung and OKI. Ammonium concentration in Bandung reached 4.94 µg/m³ (sample code: BDG2-06), and in OKI, 4.49 µg/m³ (sample code: OKI003). In Serpong, the ammonium concentration was almost similar to the nitrate ion concentration, ranging from 2.27 to 3.93 µg/m³ for ammonium and from 2.07 to 3.30 µg/m³ for nitrate ions.

Figure 6b shows that the lower PM2.5 concentrations in OKI and Serpong have higher values of total ion composition (in %) compared to Bandung, except for sample SRP 15, which was conducted at nighttime and resulted in a lower value of total ions. The reduced total ion concentration in this nighttime sample may be attributed to lower pollutant emissions during the night.

Backward trajectories (Hysplit)

The frequency distribution of HYSPLIT trajectories for pollutants from October 29, 2021, to November 7, 2021, is presented in Figure 7a. In general, 90% of the pollutants were of local origin, located within the vicinity of Bandung. Of this total, 30% consisted of pollutants transported from other regions, specifically through long-range transport from the northeast of Bandung. This suggests a hypothesis that, during this period, approximately 60% of the pollutants in the Bandung area originated from local sources, while the remaining 40% were transported from the northeast, from regions such as Cirebon and Indramayu.

Figure 7.

HYSPLIT Frequency of Backwards Trajectory in (a) Bandung in the 29 October 2021-7 November 2021 period, (b) Serpong in the 23-27 July 2022 period, (c) OKI South Sumatra in the 1-6 Oct 2021 period

However, a key point to highlight in this study is that during this period, Bandung did not experience large-scale long-range transport of pollutants. Most pollutant sources were local, with only 30% resulting from long-range transport, which was limited to the Indramayu area. This observation leads to the hypothesis that wind speeds heading toward Bandung during this period were not strong enough to carry pollutants over long distances, resulting in a predominance of locally generated pollutants.

Such findings underscore the importance of both local emission sources and regional pollutant transport in contributing to air quality in Bandung. Although long-range transport can significantly influence pollutant concentrations in urban areas, as evidenced in similar studies on transboundary air pollution (Yin et al., 2021), the current period indicates that local sources played a more dominant role.

The frequency distribution of HYSPLIT trajectories for pollutants from October 29, 2021, to November 7, 2021, in Serpong is presented in Figure 7b. In contrast to Bandung, the cumulative frequency analysis reveals that only about 70% of the pollutants in Serpong are of local origin. In comparison, 30% are the result of long-range transport, with some contaminants travelling a considerable distance from the Indian Ocean, south of East Java. This suggests that only 40% of the pollutants in Serpong are local, with the remainder attributed to long-range transport.

The significant distance covered by pollutants arriving in Serpong suggests that, during this period, strong winds were blowing toward and from Serpong. This hypothesis supports the idea that extensive long-range pollutant transport occurred, driven by high wind speeds, making long-range transport a dominant factor in Serpong's air quality during this period.

Such patterns of pollutant dispersion are crucial for understanding the dynamics of air pollution in Serpong, as long-range transport can dramatically affect pollutant concentrations, especially during periods of high wind activity. This finding aligns with broader studies on regional pollution dynamics, where wind speed and direction play a significant role in the dispersal of air pollutants across large distances (Nguyen et al., 2022).

During the period from October 1 to 6, 2021, OKI experienced a similar situation to Bandung, as shown in Figure 7c. In this period, the majority of the pollutants were attributed to local emissions, with minimal contribution from long-range transport. Based on cumulative frequency calculations, more than 90% of the contaminants identified during this period were attributed to local emissions.

This highlights a pattern where local emission sources played a dominant role in OKI's air quality during this timeframe. The minimal impact of long-range transport suggests that external factors, such as wind patterns or distant pollution sources, had a negligible effect on the air quality in OKI at this time. This phenomenon underscores the importance of controlling local emissions to improve air quality in urban areas. This conclusion has been drawn in other studies focusing on regions with low long-range pollutant transport (Guo et al., 2019). The findings suggest that external factors, such as wind patterns and the transport of pollutants from distant regions, played a crucial role in elevating pollutant levels. Such large-scale transport of pollutants is a common occurrence in the areas affected by vigorous wind activity, which can carry pollutants over long distances, as noted in studies of cross-regional pollution events (Duc et al., 2021).

Aerosol Optical Thickness (AOT) from Himawari satellite observation and association with respiratory diseases

Satellite-derived aerosol products, particularly aerosol optical thickness (AOT), have been widely used as proxies for surface PM2.5 concentrations in regions with limited ground-based monitoring, as they capture regional aerosol loading and spatial–temporal variability of particulate matter (Gupta et al., 2006; van Donkelaar et al., 2015). In this study, PM2.5 and Himawari-8 AOT exhibit partially consistent temporal patterns during several sampling periods; however, their overall association remains weak to moderate (R=0.396) as shown in Fig. 8 (a) and (b). This reflects the fundamental difference between column-integrated AOT and near-surface PM2.5, which is strongly influenced by boundary layer dynamics, local emissions, and meteorological conditions. Similar weak to moderate AOT–PM2.5 relationships (R< 0.5 and 0.5<R<0.7) have been reported in previous Himawari-based studies across East and Southeast Asia and also Indonesia (Indrawati et al., 2025; Yoshida et al., 2018; Fu et al., 2023). The limited number of paired observations in this study further reduces statistical robustness, while frequent cloud cover, high relative humidity, and aerosol hygroscopic growth in tropical regions constrain the availability and accuracy of Himawari AOT retrievals. Nevertheless, numerous studies have shown that AOT remains a valuable indicator of regional PM2.5 exposure when interpreted at appropriate spatial and temporal scales or combined with auxiliary variables, supporting its continued use as a complementary tool for air quality assessment in data-sparse regions (Kloog et al., 2012; Li et al., 2017).

Figure 8.

Pattern of the daily AOT Himawari with PM2.5 concentration in three sites during the sampling period (a) and scatter plot between AOT Himawari and PM2.5 concentration (b)

Figure 9 presents the spatial co-variation of Aerosol Optical Thickness (AOT), retrieved from Himawari satellite observations, and the prevalence of respiratory diseases in ten Indonesian provinces (including South Sumatra, West Java, and Banten, where the sampling of PM2.5 was conducted) for the years 2018 and 2023. The overall AOT values range from 0.3 to 0.6, indicating moderate to high levels of atmospheric aerosols that may impact air quality and human health.

Figure 9.

Respiratory disease prevalence associated with AOT value in all ages (a), acute respiratory tract infection (ARTI) and pneumonia prevalence in toddlers (b), and asthma relapse prevalence in all ages (c) in ten provinces in Indonesia, including South Sumatra, West Java, and Banten provinces in 2018 and 2023.

In 2018, the provinces of South Sumatra, Riau Island, and Banten exhibited relatively higher AOT values (approximately 0.5), consistent with regional haze events and biomass burning occurrences in Sumatra and parts of Java. Correspondingly, higher prevalence rates of Acute Respiratory Tract Infection (ARTI) and pneumonia were recorded in almost all provinces. For instance, ARTI prevalence exceeded 5% in Banten in 2018 and more than 3% in South Sumatra and Riau Island in the same year, suggesting a potential association between increased aerosol loading and respiratory morbidity. Similar patterns have been reported in previous studies, where elevated aerosol optical depth and particulate matter from biomass burning significantly increased respiratory disease incidence in Indonesia and neighbouring Southeast Asian regions (Crippa et al., 2016; Koplitz et al., 2016; Lelieveld et al., 2020). A similar trend was observed among toddlers, where ARTI_toddler and pneumonia_toddler prevalence reached up to 10–12% during periods of elevated AOT, aligning with findings that children are particularly susceptible to fine particulate exposure due to immature respiratory defences (Wang et al., 2019; Reddington et al., 2021).

In contrast, during 2023, AOT values showed a slight decline (ranging from 0.3 to 0.4) across all provinces, accompanied by a decrease in ARTI and pneumonia prevalence in both general and toddler populations in almost all provinces. This reduction may be attributed to improved air quality and strengthened air pollution control policies implemented during the intervening years, including the expansion of emission monitoring and fire prevention initiatives (Ministry of Environment and Forestry Indonesia, 2022; UNEP, 2023). Nevertheless, asthma relapse prevalence remained consistently high (50–70%) regardless of AOT variations, indicating that asthma exacerbations are likely influenced by multiple factors, including meteorological conditions (humidity, temperature), allergen exposure, and genetic predisposition (Zhang et al., 2020; Zhang et al., 2024), rather than solely by aerosol concentration.

Overall, the results indicate a positive relationship between AOT and the prevalence of acute respiratory diseases. Elevated aerosol loading, as reflected by high AOT values, may enhance particulate matter concentration (PM2.5 and PM10), which has been widely documented to aggravate respiratory symptoms and reduce lung function (Yan et al., 2025; Yee et al., 2021; WHO, 2021). These findings are consistent with previous studies that have linked satellite-derived aerosol metrics to population-level respiratory outcomes in Southeast Asia, particularly during biomass burning seasons.

Hence, the integration of satellite-based aerosol monitoring and public health surveillance provides a valuable early-warning approach for detecting and mitigating respiratory health risks in regions vulnerable to haze and air pollution. Continuous monitoring using satellite products, combined with ground-based health data from local health agencies, can further enhance understanding of the spatiotemporal dynamics between aerosol exposure and respiratory diseases in Indonesia. The consistent positive pattern across all disease categories reinforces the potential role of AOT as an early indicator of air-quality-related respiratory health risks. Similar findings have been reported in Southeast Asia, where satellite-derived aerosol data were shown to effectively capture spatial and temporal variations in respiratory disease incidence linked to haze and biomass burning events (Crippa et al., 2016; Reddington et al., 2021; UNEP, 2023). These results reinforce the descriptive findings presented earlier, supporting the hypothesis that elevated aerosol optical thickness, as a proxy for fine particulate pollution, is strongly linked with increased respiratory disease prevalence (Crippa et al., 2016; Mo et al., 2018; Reddington et al., 2021).

A Pearson correlation analysis was performed to evaluate the relationship between Aerosol Optical Thickness (AOT) and the prevalence of respiratory diseases across the ten provincial-year observations (including South Sumatra, West Java, and Banten for 2018 and 2023). As shown in Table 3.1, all the respiratory diseases exhibited weak to moderate positive correlations with AOT. This suggests that higher aerosol concentrations—reflected by increased AOT—are possibly associated with incidences of acute respiratory infections.

Table 3.

Correlation analysis of AOT and respiratory diseases

The weak to moderate positive correlations between AOT and ARTI (R=0.19), pneumonia (R=0.30), and asthma (R=0.34) are consistent with the co-spatial variation explained in Fig. 9, where provinces with higher AOT values generally exhibit higher prevalence of these respiratory outcomes, suggesting that exposure to particulate air, especially during haze events likely exacerbates both acute and chronic respiratory outcomes. This finding aligns with previous studies indicating that satellite-derived AOT can serve as a proxy for regional aerosol exposure associated with increased respiratory morbidity, particularly for ARTI and asthma, which are sensitive to short- and medium-term particulate exposure (Gupta et al., 2006; van Donkelaar et al., 2015; Guarnieri & Balmes, 2014). The consistency of the pattern across provinces and years underscores the potential of AOT as a satellite-based indicator for assessing respiratory health risks in Indonesia.

Weak positive correlation was observed for TBC (R = 0.2), reinforcing the understanding that tuberculosis is predominantly driven by socioeconomic conditions, population density, and access to medical services rather than ambient aerosol levels alone (Brook et al., 2010). Asthma prevalence and AOT showed only a weak to moderate correlation (R = 0.34), implying that asthma is more likely influenced by multiple environmental and physiological factors beyond aerosol exposure, such as humidity, allergens, and genetic predisposition (Zhang et al., 2020; Liu et al., 2022). The negative correlation between AOT and asthma relapse (R = –0.31) is consistent with the heterogeneous pattern observed in Figure 9(c), where higher AOT does not systematically correspond to higher relapse prevalence. This suggests that asthma relapse is influenced by a complex interplay of factors, including disease management, medication adherence, and temporal exposure variability, which are not fully captured by province-level AOT indicators (Guarnieri & Balmes, 2014; Varopichetsan et al., 2025). From this study was known that AOT is a useful ecological indicator of regional aerosol loading associated with respiratory disease prevalence. However, the generally modest correlation coefficients highlight the multifactorial nature of respiratory diseases and the limitations of ecological analyses in inferring individual-level health effects.