Introduction

Surface ozone (O3), a secondary pollutant from complex photochemical reactions between precursors and sunlight, adversely affects human health and vegetation1,2. Changes in photochemical activity and meteorological parameters can alter surface O3 concentrations3,4,5. Elevated temperatures and solar radiation accelerate the production of O36. Stagnant conditions and lower planetary boundary layers have also been linked with high O3 7. Surface O3 formation is a nonlinear process involving nitrogen oxides (NOx) and volatile organic compounds (VOCs) as precursors8,9. Thus, precursor emission changes can also modulate O3 concentration in addition to meteorological factors. NOx is primarily produced from anthropogenic sources, whereas VOCs arise from both natural (biogenic) and anthropogenic sources. O3 sensitivity to the precursors depends on the photochemical regime of the O3 formation10. Photochemical O3 formation is influenced by a ratio of VOC to NOx. The formation regimes are generally classified into VOC-limited regime (where O3 formation is sensitive to an increase in VOCs), NOx-limited regime (where O3 concentration is mainly affected by an increase in NOx and is insensitive to VOCs), and can be in a transitional state4,11,12,13. Understanding O3 formation sensitivity is essential for establishing practical O3 abatement efforts and remains a scientific challenge that must be addressed. Formaldehyde (HCHO) to nitrogen dioxide (NO2) Ratio (FNR), derived from satellite measurements, serves as a valuable indicator for inferring O3 formation due to its brief lifespan in the boundary layer and close association with O3 formation14,15,16. Martin et al.14 and Duncan et al.15 combined satellite measurements and air quality models to establish the FNR threshold and concluded that FNR < 1 indicates a VOC-limited regime, FNR > 2 indicates a NOx-limited regime, and 1 > FNR > 2 indicates a transitional regime in the United States. Jin et al.17 recently derived the FNR threshold for major U.S. cities with higher FNR values (3–4) in the transitional regime. In addition, Wang et al.18 revealed that a VOC-limited regime occurs when FNR < 2.3, a NOx-limited regime when FNR > 4.2, and an FNR between 2.3 and 4.2 reflects the transition regime in major Chinese cities.

Extensive research using modeling, ground measurements, and satellites has delved into surface O3 trends and causes, and its formation sensitivity, yet Southeast Asia, particularly Indonesia, remains inadequately explored, as evidenced by limited studies19,20,21. The scarcity of O3 measurements and a predominant focus on particulate matter reduction in Indonesia likely contribute to this research gap22. Indonesia currently installs 56 Air Quality Monitoring Stations (AQMS), twelve located in the Jakarta Greater Area (JGA). According to Verma et al.23, considering the vast population of 273.8 million, the number of AQMS in Indonesia is still very few and requires 1039 more stations. JGA is Indonesia’ s largest urban agglomeration, comprising Jakarta’ s capital and satellite cities such as Bogor, Depok, Bekasi, and Tangerang. The JGA has a dense population, rapid economic development, intensive industrialization, and high transportation utility. Bogor, located in the southern part of Jakarta, encompasses an area five times larger than that of Jakarta. With distinctive topography, Bogor functions as a downwind location influenced by local and transported emissions. An air quality station in Cibeureum (CBR), southeast Bogor, reflects a background environment close to Jakarta’ s.

This article unveils the seasonal variations and temporal trends in O3 concentrations within urban Jakarta and rural Bogor employing ozone metrics. The investigation assesses the contribution of meteorological and anthropogenic factors to O3 changes utilizing stepwise Multiple Linear Regression (MLR). Furthermore, the study analyzes O3 formation sensitivity using FNR derived from satellite observations and compares them with the previous studies. This study extends beyond Jakarta, offering insights with broader implications for other megalopolitan areas grappling with similar environmental challenges.

Results

Seasonal variation and temporal trends of ozone (O3)

Figure 1 depicts the time series of maximum daily 8-h average (MDA8) monthly anomalies and the trends in the 10th, 50th, and 90th percentiles at five sites in urban Jakarta (hereinafter refers to DKI) and one rural site of CBR. All DKI sites generally exhibit a downward trend of surface O3 in the 50th percentile, as presented in Table 1. Highest O3 declining rate occurs in DKI1, DKI2, DKI4, DKI5, and DKI3 by about − 8.0 µg m−3 year−1, − 7.1 µg m−3 year−1, − 5.7 µg m−3 year−1, − 0.3 µg m−3 year−1, and − 0.2 µg m−3 year−1, respectively. The tenth percentile represents low O3 concentration that may not involve or is less involved in photochemical reaction24. DKI2 in northern Jakarta appeared to have the highest and significant declining rate of − 6.4 m−3 year−1 for the 10th percentile, followed by DKI4 (− 4.6 µg m−3 year−1), DKI5 (− 3.3 µg m−3 year−1), DKI3 (− 3.1 µg m−3 year−1), and DKI1 (− 2.8 µg m−3 year−1).

Figure 1
figure 1

Map of the study area and the time series of the monthly mean anomaly of ozone maximum daily 8-h average (MDA8 O3) during the study period at Jakarta and Bogor sites. Color solid lines denote linear trends in the 90th (blue), 50th (red), and 10th (green) percentiles. Pink shade represents Jakarta Province with sites in central Jakarta (DKI1), north (DKI2), south (DKI3), east (DKI4), and west (DKI5). CBR represents the site in rural Bogor. The shapefile for Jakarta Greater Area boundary map is obtained from Geospatial Information Agency of Indonesia https://geoservices.big.go.id/petarbi/ and the map plot is generated using ArcGIS Pro online 2.8.0. The plot graph is generated using RStudio 1.4.1106. 2021.

Full size image
Table 1 Slope values from QR annual trend analysis for different percentiles at DKI (2010–2019) and CBR sites (2017–2019).
Full size table

Meanwhile, MDA8 O3 anomaly trends in the 90th percentile show different directions from one site to another. Central (DKI1), south (DKI3), and east (DKI4) Jakarta shows a decreasing trend in different order of magnitude. DKI1 and DKI4 experience faster concentration decreases and significant rates of − 10.7 µg m−3 year−1 and − 7.7 µg m−3 year−1. DKI1 and DKI4 perform at faster-decreasing rates than their lower (50th and 10th) percentiles, indicating reduced extreme ozone episodes during the study period. Despite the reduction rate of O3 in the 50th and 10th percentiles, DKI2 and DKI5 present insignificant upward trends in the 90th percentile with a rate of 1.9 µg m−3 year−1 and 5.9 µg m−3 year−1, respectively, suggesting increase number of extreme O3 concentration.

The MDA8 O3 (average from all DKI sites) exceedance frequency (> 100 µg m−3 for 8-h threshold) in the dry season showed a sharp rise after 2010, fluctuated until 2018, but underwent a remarkable reduction in 2019, as demonstrated in Supplementary Fig. S1a. The minimum (maximum) exceedance frequency was observed in 2019 (2012). A yearly average of MDA8 O3 concentration surpassing the 8-h threshold, calculated from daily data, shows a slight reduction from 2011 to 2019 (see Supplementary Fig. S1b). This finding corresponds with the significant downward trend in high O3 concentration (90th percentile) observed at specific DKI sites, as depicted in Fig. 1 and Table 1.

The MDA8 O3 increased during the dry period across DKI and CBR sites, peaking in October. Figure 2 displays a map of the average maximum daily 8-h of ozone (AVGMDA8) in urban Jakarta and rural Bogor during the study period for the dry season (April to November). The AVGMDA8 varied and showed consistency with the 1-h O3 value, with the highest concentration at DKI3, followed by DKI5, DKI2, DKI4, and DKI1 with values of 128, 127, 122, 111, and 99 µg m−3, respectively. During 10-year period from 2010 to 2019, all DKI sites exceeded the National Ambient Air Quality Standard (NAAQS) for 1-h concentration of O3 (1-h O3 >150 µg m−3) with 18%, 26%, 30%, 37%, and 41% of exceedances, respectively, at DKI1, DKI4, DKI2, DKI3, and DKI5, as presented in Supplementary Fig. S2. The CBR experienced the 1-h O3 concentration threshold exceedances of 0.1% during 2017 to 2019. The monthly variations in 1-h O3 at the DKI and CBR sites peaked between May and November with an upward trend. The O3 concentrations gradually decreased from December to April. During the observation period, DKI5 experienced the highest annual average of 1-h O3 concentration compared to other sites with a maximum (minimum) of 194 µg m−3 (97 µg m−3), followed by DKI2 and DKI3, which were 160 µg m−3 (107 µg m−3) and 147 µg m−3 (103 µg m−3), respectively.

Figure 2
figure 2

Map of the average of Maximum Daily 8-h ozone (AVGMDA8) in urban (Jakarta) and rural (Bogor) during the dry season (April to November) over the entire study period. (ArcGIS Online Pro 2.8.0. 2021). The shapefile for Jakarta Greater Area boundary map is obtained from Geospatial Information Agency of Indonesia https://geoservices.big.go.id/petarbi/ and the map plot is generated using ArcGIS Pro online 2.8.0.

Full size image

The overall O3 level at the CBR site was considered low compared to that in DKI sites and the NAAQS; however, the annual 1-h and MDA8 values increased from 2017 to 2019. MDA8 O3 anomaly increased in all percentiles by approximately 1.3 µg m−3 year−1 (90th), 4.2 µg m−3 year−1 (50th), and 3.4 µg m−3 year−1 (10th), indicating the enhanced presence of precursors that have critical roles in photochemical O3 formation in this area. Changing O3 concentration in the 10th percentile represents changes in the local background of O3 concentration24.

Meteorological impact on the MDA8 O3 variation

The MLR model yielded the adjusted coefficients of determination (R2) for Jakarta (average from five DKI sites) and CBR areas for all seasons of 0.28 and 0.43, respectively, implying that 28% and 43% of the MDA8 O3 daily variability were associated with meteorological conditions. The contributions of meteorological and anthropogenic factors to the monthly mean MDA8 O3 anomalies in Jakarta and CBR areas over the entire period are depicted in Fig. 3a,b, respectively. Meteorological components influenced inter-annual O3 changes in Jakarta, accounting for 30% of the relative contribution for the entire season of 2013–2019. In general, meteorological components contribution ranged from 0.2 to 98.5% and was not considerably different from that of the dry season (Supplementary Fig. S3a). A slightly lesser meteorological influence shows during the wet season with a 25% relative contribution to O3 variability.

Figure 3
figure 3

Contribution of meteorological (green bar) and anthropogenic factor (blue bar) to observed MDA8 O3 anomalies (black line) in (a) Jakarta (2013–2019) and (b) Bogor (2017–2019). (Matlab R2014b, 2014. 416,517).

Full size image

Pronounced positive contributions from meteorological features such as February 2018 altered MDA8obs very slightly. Anthropogenic factors contributed 70% to the inter-annual variability of ΔMDA8obs, averaging 1.5–99.8%. Notably, positive ΔMDA8obs in 2013, half of 2014, and 2018 were primarily attributed to anthropogenic rather than meteorological. The positive anthropogenic contribution reached more than 90% and was responsible for the increase of monthly mean ΔMDA8obs during those years. The monthly mean of ΔMDA8obs fluctuated and increased significantly in 2018 (Fig. 3a). These results indicate that changes in anthropogenic emissions are more favorable for O3 production, particularly during the dry season.

Downward UV radiation at the surface (UVB) significantly contributed to the O3 variation in all seasons, although the coefficient was minimal (Table 2). UVB also revealed a positive correlation with the correlation coefficient of r = 0.32, the second-highest value after wind speed (Supplementary Table S2). The UVB pattern was synchronous with maximum temperature (Tmax), particularly during the dry season, as shown in Supplementary Fig. S4c,d. Intense UV radiation is favorable for ozone production. Furthermore, higher temperatures can increase HCHO concentrations from biogenic emissions25. More than 70% of the daily observed MDA8 was above the NAAQS and coincided with UVB > 70,000 J m−2. According to Indonesian Meteorological Office (BMKG), the UV index in Jakarta at 10:00 local time was estimated to be 2–6, indicating a moderate to high risk. A negative anomaly of UVB influenced the O3 decline in 2013 (− 8.5 µg m−3) (Supplementary Fig. S4d). However, the high negative anomaly of UVB in 2016 (− 7759 J m−2) was likely canceled out by the positive contribution of Relative Humidity (RH), thus inducing only a small perturbation in O3.

Table 2 Coefficient of several variables from the multiple linear regression (MLR) model results for Jakarta (2013–2019) and CBR sites (2017–2019) in the dry, wet, and all seasons.
Full size table

UVB and wind speed mean (WS_MEAN) showed higher correlations than other parameters in Jakarta. The wind speed significantly (p = 2 × 10−16) contributed to the stepwise MLR. Wind speed is negatively correlated with MDA8 O3. During the dry period, air stagnation frequently occurred in Jakarta, boundary layer height (BLH) was low, calm winds dominated and recorded as approximately 70% of prevailing winds, and the wind speed was < 3 m s−1 leading to pollutant accumulation on the surface26,27. There is likely a contribution from the significant negative anomalies of the mean wind speed in 2018 (− 0.2 m s−1) to the increased MDA8met (5.9 µg m−3) during the dry period (in Supplementary Fig. S3a). The BLH contributed negatively to Jakarta’ s MDA8 O3 change, indicating that shallow BLH increases O3 concentration. In addition, the immense positive contribution from emission drivers in 2018 substantially influenced the overall increase in O3 levels at the DKI sites. High RH is associated with enhanced cloud cover and a greater chance of rainfall, slowing the photochemical reaction of ozone formation3,4.

Meteorological and anthropogenic components shared almost equal proportions over 3 years’  period of measurement from 2017 to 2019 at CBR site, with average proportions of 52% (ranging from 17 to 91%) and 48% (ranging from 9 to 83%), respectively. However, the number did not change significantly during the dry season. UVB (p = 3 × 10−6), BLH (p < 2 × 10−16), Tmax (p = 0.005), and Surface Pressure (SP) (p = 1.7 ×10−6) significantly contributed to O3 variation at CBR. The CBR site exhibits complex terrain and thus presents unique weather conditions. High-altitude areas experience lower pressure. With a forced inversion layer and slow wind speeds, vertical mixing and horizontal dispersion in the atmosphere are poor and hindered by mountains, leading to increased concentrations of pollutants28. BLH was significant in CBR and had moderate positive correlations. In higher-altitude areas, during the development of a boundary layer, intrusion of O3 from the upper atmosphere to near the surface is possible, thus contributing to the surface O3 level29,30. This could be a possible reason for a positive correlation between BLH and O3 concentration. A similar characteristic of BLH and O3 variations observed at the CBR site is also evident in other locations with comparable geographic features. For instance, Mount Lulin in Taiwan31 and certain sites in in Beijing-Tianjin-Hebei Province and the Yangtze River Delta, China4.

Spatial and temporal analysis of OMI NO2 and HCHO column density

The NO2 monthly average in Jakarta ranged from 1.5 to 8.5 × 1015 molecules cm−2, with a 10-year average of 3.6 × 1015 molecules cm−2 (Fig. 5a). This value is lower than those of cities in Japan13, China9,11, London32, and Mexico city12 for the same study period. The trend in the 50th percentile depicted an insignificant downward trend, with rates of − 0.04 × 1015 and − 0.01 × 1015 molecule cm−2 year−1 in Jakarta and CBR, respectively. The tropospheric NO2 in CBR is lower than in Jakarta, with a 10-year average of 2.9 × 1015 molecule cm−2 and ranging from 1.1 to 5.5 × 1015 molecule cm−2. Although the trend slightly decreased, a substantial increase was observed in the NO2 OMI from 2017 to 2019, with a slope of 0.02 × 1015 molecules cm−2 month−1.

HCHO column density presented an insignificant upward trend of 0.19 × 1015 (0.23 × 1015) molecule cm−2 in Jakarta (CBR). 10-year average of OMI HCHO column density in urban Jakarta was 12.6 × 1015 molecules cm−2 with monthly average ranging from 0.9 to 19.4 × 1015 molecules cm−2 (Figs. 4a and 5b). This number is greater than that of urban areas in China (~2 to 18 × 1015 molecules cm−2 in Beijing and Shanghai), Japan (~1 to 16 × 1015 molecules cm−2 in Tokyo, Osaka, and Nagoya), and Mexico City (~3 to 9 × 1015 molecules cm−2)12,33.

Figure 4
figure 4

Monthly average of (a) HCHO column density and (b) NO2 tropospheric column from 2010–2019 in Jakarta and Cibeureum Bogor. The dashed line represents a trend in the 50th percentile. (RStudio 1.4.1106. 2021).

Full size image
Figure 5
figure 5

Map of the average (a) OMI NO2 tropospheric column, (b) OMI HCHO vertical column, and (c) FNR value from 2010–2019 for all seasons; (d), (e), and (f) for dry seasons; and (g), (h), and (i) for the wet season. Transparent areas indicate an FNR value above 5. The shapefile for Jakarta Greater Area boundary map is obtained from Geospatial Information Agency of Indonesia https://geoservices.big.go.id/petarbi/ and the map plot is generated using ArcGIS Pro online 2.8.0.

Full size image

OMI HCHO in CBR area is lower than in Jakarta, with a yearly average of 10.9 × 1015 molecules cm−2 over the entire study period (Fig. 5b). Biogenic VOC emissions from plants and vegetation are more reactive than anthropogenic VOCs, causing relatively high HCHO columns over cropland areas such as CBR. However, anthropogenic Non-Methane VOC (NMVOC) emissions surpass biogenic VOC emissions11 in urban cities where anthropogenic activities are massive and intense. Anthropogenic NMVOC emissions in Bogor from on-road motorized vehicles was around 9.7 Gg year−1 in 201634, one-fifth smaller than Jakarta’ s emission for the same sector (48.6 Gg year−1 in 2015)35.

Seasonal variation of OMI NO2 and HCHO

Seasonal variations in NO2 concentrations in Jakarta and CBR exhibited the same pattern, increasing in April and gradually decreasing in October. In Jakarta, the maximum (minimum) NO2 concentration occurred during the dry (wet) period in July (December) (Fig. 5d,g, and Supplementary Fig. S7). Enhanced column density occurred in northern Jakarta and extended to Bogor city center during the dry period (Fig. 5d). We also examine the monthly variability of NO2 from in-situ measurements and reveal that the concentration fluctuates throughout the year, and the seasonal pattern is not too obvious, as depicted in Supplementary Fig. S7. Sofyan et al.36 found no significant differences in fuel consumption in Jakarta between the dry and wet seasons. However, there is an apparent effect of sea breeze on transporting NO2 from north Jakarta to central and southern Jakarta. On dry days, converging sea breezes in the southern part of Jakarta cause pollutant accumulation in central Jakarta36. The pollutants then move along the sea breeze front and penetrate the Bogor area.

Generally, NO2 in Jakarta was higher than in CBR for the whole seasons. The maximum NO2 concentration in CBR occurred in September, and the minimum in December. The reduction in NO2 was mainly due to meteorological conditions during the wet period. Disregarding exceptional circumstances such as movement restrictions during the COVID-19 pandemic, which were not considered in this study, anthropogenic emissions in Jakarta and CBR were steady throughout the year.

Compared to NO2, HCHO was higher during the dry than wet period but fluctuated more (Fig. 5e,h). A slightly higher temperature during the dry period accelerated VOC photochemical oxidation, thereby contributing to high levels of HCHO. Elevated HCHO levels were notable and extended northwest and northeast of Jakarta during the dry period from 2010 to 2019, as shown in Fig. 5e. The HCHO column density in the whole Bogor area varied significantly seasonally. However, it was slightly lower in northwestern Bogor than in central Bogor.

OMI FNR

Figure 5c presents the spatial yearly average FNR value from 2010 to 2019 in JGA. FNR in Jakarta steadily increased to 4.5, consistent with the increase in HCHO column density (Fig. 4a). Seasonal variations in the FNR occurred in Jakarta and CBR (Fig. 5f,i, and Supplementary Fig. S9b). As expected, the FNR decreased in Jakarta’ s dry period but fluctuated slightly in CBR. The reduction in FNR during the dry period was consistent with that shown in Supplementary Fig. S7, where the NO2 tropospheric column increased.

During the dry period, the maximum 1-h O3 and MDA8 O3 showed more days exceeding the 1-h and 8-h NAAQS thresholds than the wet period. It is worth noting that a large positive MDA8 O3 anomaly in 2018 was mainly driven by a large positive anomaly in anthropogenic activity. The increased MDA8 O3 is likely due to a response from the increase of OMI NO2 under a high FNR value of 3.9. However, to confirm this finding, further modeling work should be conducted.

Discussion

O3 concentration and its formation mechanism are influenced by meteorology and changes in VOC and NOx. The O3 formation sensitivity could be either VOC-sensitive, NOX-sensitive, or transitional, where an increase of both precursors affects the O3 production. In a vibrant urban agglomeration such as JGA, characterized by intense emissions and limited and scattered AQMS, investigating the long-term trends of O3 pollutants and their formation regimes becomes imperative for effective air quality management. Identification of O3 formation sensitivity aids policymakers in designing appropriate control strategies. The current study evaluates trend and meteorological influences on O3 concentration variability. Additionally, we leverage space-borne NO2 and HCHO tropospheric column density to compute FNR as a VOC/NOX ratio proxy to characterize O3 formation regimes.

Generally, our results show a downward trend of MDA8 O3 anomalies at all sites except CBR of rural Bogor. Despite the reduction trend, some sites experienced an increasing trend of high O3 episodes, such as in northern and western Jakarta, and almost steady state in southern Jakarta. This feature leads to high episodes exceeding the national standard for 1-h and 8-h O3 concentration from 2010 to 2019. As much as 41% of 1-h O3 concentration exceeded the national threshold for the whole seasons from 2010 to 2019 and 93% for MDA8 O3 in the 2018 dry season for Jakarta, with annual averages reaching 158.9 µg m−3. It is worth noting that high positive anomalies of O3 during 2018 are influenced mainly by anthropogenic drivers rather than the meteorology, as can be seen in Fig. 3a. It is likely due to a response from the increase of OMI NO2 under increase FNR value of 3.9 (Fig. 4b). However, to confirm this finding, further modeling work should be conducted.

DKI1 is in the city center, while DKI2 is near the busy harbor, power plants, and industrial areas. Both sites are characterized by higher NOX emissions owing to high anthropogenic activities, as reflected by higher NO2 concentration (see Supplementary Fig. S8). In the vicinity of high NO emissions, O3 is removed via NO titration. However, its concentration increases further downstream, as in DKI3 and DKI5. DKI3 and DKI5 presented higher exceedance episodes and the least declining O3 trend. During the dry period, wind direction toward Jakarta mainly originated from the east and southeast, allowing the transport of pollutants from outside Jakarta to the eastern part of Jakarta, where industrial and manufacturing centers are located (around the Bekasi industrial area), contributing to the increase in O3 concentration at DKI5 (western Jakarta). In July, the HYbrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model’ s backward trajectory showed that air mass transport originated from Karawang and passed through Bekasi and South Jakarta (Supplementary Fig. S10). Despite the synoptic scale of easterly and southeasterly winds during the dry season, Jakarta was affected by sea breezes. According to Kitada et al.37, the southerly land breeze carried pollutant-rich but ozone-poor air masses to the Jakarta coast at night. In the morning, photochemical reactions produce O3 in northern Jakarta, and sea breeze circulation develops in the lower layers under weak synoptic winds37. The sea breeze, opposite the easterly and southeasterly winds, became dominant owing to the easterly and southeasterly synoptic scales blocked below approximately 1 km by the mountains in the southern part of West Java. The O3-rich air mass was transported to southern Jakarta and Bogor as the sea breeze propagated inland up to 60 km from the Jakarta coastline37,38. This sea breeze phenomenon affects the convection activity and transport of atmospheric pollutants, which explains the high concentration of DKI3 downwind of Jakarta.

Measurement of OMI NO2 in Jakarta shows an insignificant decreasing trend and increasing HCHO from time to time. In metropolitan cities in the US, China, Mexico, and Japan, tropospheric NO2 outnumbers the HCHO. Thus, the O3 formation is VOC-limited regime, as shown in Table S5. However, that is not the case for Jakarta, where HCHO column leads. As a consequence, the FNR falls in higher value compared to those cities. Study from Lestari et al.35 shows that the most significant emission load was from CO, NOX, and NMVOC, at approximately 52.9, 143.9, and 48.6 kilotonnes in 2015, respectively. Significant contributors to NOX were the road transportation sector (57%), power plants (24%), industrial sectors (15%), and residential areas (4%)35,39,40. Furthermore, by Lestari et al.35, the total emissions inventory of NMVOC in Jakarta was comparable to that of NOX in the same sectors and years. Road transport was the major contributor to NMVOC in Jakarta, with a percentage of 96%, followed by industry at 2%, power plants and residential areas at 1%. According to the Ministry of Transportation of the Republic of Indonesia, there was a total transportation movement of 19.63 million movements/day in 2022. Road transport by motorcycles contributed to 88% of the total NMVOC emissions. Jakarta Provincial Agency documented that there were over 15.8 million motorcycles in the city in 2019, approximately a 2.7% increase from 2018. Motorcycles represent the dominant vehicle fleet in Jakarta Province, accounting for approximately 80% of the total vehicles, followed by cars (17%), trucks (3%), and buses (0.2%)41. Similar emission types were also found in Vietnam, where motorcycles mainly contributed to the VOC emissions in Hanoi and Ho Chi Minh City. Motorcycles comprise 80% of the total transportation in Vietnam, and the number of vehicles per 1000 population in Vietnam is lower than that in Jakarta42,43. Emissions from motorcycles are likely attributable to high HCHO column density in Jakarta. In addition, NMVOCs are abundant in certain areas, such as gas stations44 and solvent usage industries45.

Different features occurred at the rural CBR site. A 3-year measurement period shows that the O3 concentration is below the government threshold; however, it shows an increasing rate for all percentiles. The growing O3 concentration value at the rural site indicates the enhanced presence of precursors that have critical roles in photochemical O3 formation. Changes in O3 concentration in the 10th percentile represent changes in the local background of O3 concentration. Unlike Bogor, a decreasing trend in O3 was observed in the rural North China Plain, Europe, and North America, likely because of the implementation of the O3 control policy24.

Bogor has a lower population density and more vegetation than Jakarta. The number of vehicles in Bogor is two million, one-thirteenth of those in Jakarta (26 million)41. Bogor, as the outskirt of Jakarta, may receive pollutants transported from Jakarta along with the sea breeze; however, this may not contribute substantially to the total emissions of the CBR. Lower concentrations recorded at the CBR were expected owing to lower emission activity, local characteristics, and meteorological disparity.

Air pollution prevention policies have been implemented in Indonesia and locally in JGA for more than a decade. For example, national kerosene-to-LPG conversion for cooking has been proven to reduce emissions40. Decreasing particulate matter, sulfate, and nitrate concentrations in Jakarta since 2006 have been used as evidence for the successful implementation of national and local policies, such as Regulation No. 141 concerning the Utilization of Gas Fuel for Public Transportation and Local Government Operational Vehicles46. Another successful implementation policy was issuing the Ministry of Environment and Forestry decree number 20/2017 on the usage of EURO IV for gasoline vehicles in late 2018. One current policy is that of the Governor of the DKI Jakarta, Regulation No. 66/2020 on vehicle emission testing, which regulates penalties for vehicles that violate the regulations.

Despite its ongoing implementation of measures, O3 pollution episodes persist. A more aggressive NOX reduction should be in place in Jakarta. Thorough investigations of the precursors and emission inventory calculation are essential to enhance our understanding of O3 formation. Moreover, the thresholds for regime classification in Jakarta may require adjustment compared to global FNR thresholds. Further researches, combining modeling and observations, are warranted to identify effective O3 reduction scenarios for informed policy recommendations.

Summary and conclusion

This study investigated variations and temporal trends in surface O3 under two different background conditions: urban (Jakarta, 2010–2019) and rural (Bogor, 2017–2019). We assessed the contributions of meteorological and anthropogenic factors and the long-term O3 formation sensitivity derived from OMI measurements. Despite the downward trend of O3 anomalies attributed to a slight reduction of NO2 concentration, Jakarta experiences recurrent elevated O3 level surpassing both the 1-h and 8-h national thresholds. Generally, anthropogenic drivers are more dominant than meteorological, irrespective of seasonal variations. However, O3 pollution can be even more severe under changing climate. Further research will focus in modeling simulations to validate the O3 sensitivity findings presented in this paper and eventually to establish O3 chemical regime thresholds, which are crucial for devising effective strategies to mitigate O3 pollution in JGA.

Data description and methodology

Study sites and air quality data

Jakarta, the center of government, business, and economic activities, contributes the most to the national gross domestic product with over 190 million USD in 201947. We used air quality data from five AQMS in Jakarta (2010–2019) operated by the Environmental Agency of Jakarta Province, one station in Cibeureum Bogor (2017–2019) operated by the Meteorological, Climatological, and Geophysical Agency (BMKG) in collaboration with the National Institute of Environmental Studies (NIES), Japan, as presented in Fig. 1 and Supplementary Table S1. Further local characteristics of Jakarta and Bogor are described in Supplementary Table S6. We used hourly O3 concentrations and preprocessed the data to eliminate negative values and outliers. The time series of the hourly concentrations were normalized using the z-score following the methodology used by Ren et al. and Sun et al.24,48. Points with an absolute z-score greater than 4 (|zt|> 4) were removed from the time series. The daily average was calculated when a station had more than 60% hourly data, and we considered daily data missing if the availability was below this threshold.

Surface O3 measurement

Jakarta Environmental Agency continuously measures 30-min surface O3 using HORIBA APOA-370 analyzer. The APOA-370 continuously monitors atmospheric ozone concentrations using a cross flow modulated ultraviolet absorption method or Non-dispersive ultraviolet absorption (NDUV). The NDUV measures ozone O3 concentration in sample gases using UV light absorption. UV light is emitted into a gas cell where the sample gas flows, and O3 absorbs UV light proportionally to its concentration. The absorbed light is detected by a photodiode, generating electrical signals. NDUV employs a cross-modulation method with a solenoid valve unit to switch between sample and reference gases, allowing the detector to distinguish AC and DC signals during processing. AC signals represent gas concentration, while DC signals compensate for UV light source aging, ensuring analyzer stability. CBR site in collaboration with NIES measure O3 concentration continuously using Kimoto OA-78749.

Meteorological data

Meteorological data were used to construct the stepwise MLR model. Parameters such as Tmax, rainfall (RRR), WS_mean, and RH of Jakarta and Bogor were obtained from the BMKG. Other meteorological parameters, such as the BLH, SP, and UVB, were downloaded from the European Center for Medium-Range Weather Forecasts Reanalysis (ECMWF ERA5) product on single levels with a spatial resolution of 0.25° × 0.25° and a temporal resolution of 1 h for the corresponding grid, similar to the meteorological stations and AQMS (https://cds.climate.copernicus.eu/cdsapp#;/dataset/).

Satellite product

We use level three gridded data of the daily ozone-monitoring instrument (OMI)50 NO2 tropospheric column standard product (OMNO2d_003) with a cloud fraction of less than 30% and a spatial resolution of 0.25° × 0.25° downloaded from https://disc.gsfc.nasa.gov/. To calculate the ratio, we average the NO2 daily data to monthly. The monthly level three vertical column density of OMI HCHO was downloaded from the European Quality Assurance for Essential Climate Variables project (QA4ECV) (http://www.qa4ecv.eu) with a spatial resolution of 0.05° × 0.05°. The QA4ECV HCHO slant column densities (SCDs) have 8–12 × 1015 molecule cm−2 uncertainties. We re-grid the monthly OMI HCHO data to 0.25° × 0.25° to match the monthly OMI NO2 data.

Trend analysis of ozone metrics

To identify O3 pollution, we use an indicator from the Indonesian Government Regulation No. 22 Year 2021 on the implementation of environmental protection and management for 1-h O3 concentration (measured from 11:00 to 14:00 local time) and O3 metric following the Tropospheric Ozone Assessment Report (TOAR), such as the maximum daily 8-h average (MDA8)51. This metric is intended to evaluate the impact of O3 on human health, vegetation, model comparisons, and the characterization of O3 in the free troposphere51. MDA8 O3 was calculated from 24 running 8-h averages. The 8-h running mean for a particular hour is computed based on the concentration for that hour plus the following 7 h. The daily maximum 8-h concentration for a given calendar day was the highest of the 8-h average concentrations computed for 8-h periods starting from that day. The 8-h running mean was considered valid when at least 5 h of data were available (60% completeness was required). The average MDA8 (AVGMDA8) is the mean value of daily MDA8 (calculated based on more than 60% of the daily MDA8).

Quantile Regression (QR) estimates the trend in MDA8 O3 concentration. QR is a well-suited technique for detecting heterogeneous distributional changes, which is often the case for free tropospheric and surface ozone as they typically show diverse percentile trends52. Trend uncertainty is estimated by doing a moving block bootstrap algorithm to take autocorrelation into account in the trend uncertainty. Trend analysis considers the seasonality of MDA8 O3 time series by calculating the MDA8 O3 anomaly. MDA8 monthly anomalies are more accurate than the monthly average in case of missing data10,48,53. Monthly anomalies were calculated by subtracting the individual monthly mean from the monthly mean for the same month for the entire study period. Meanwhile, zero anomaly refers to a situation where the monthly MDA8 O3 is equal to the long-term average value or no deviation or departure from the average conditions. QR analysis is computed using Rstudio with a quantreg package.

Stepwise multiple linear regression (MLR) model

Statistical models were used to infer the contributions of meteorological and anthropogenic factors to the monthly ozone variability. Researchers commonly use the MLR to quantify the effects of meteorological factors on O34,6,9,10,24. The MLR uses the following equation to reflect the relationship between a quantitative dependent variable and two or more independent variables:

$${{\varvec{y}}}_{{\varvec{s}},\boldsymbol{ }{\varvec{c}}}={{\varvec{\beta}}}_{{\varvec{o}}}+{\sum }_{{\varvec{k}}=1}^{{\varvec{n}}}{{\varvec{\beta}}}_{{\varvec{s}},\boldsymbol{ }\boldsymbol{ }{\varvec{c}}}\boldsymbol{ }{\varvec{x}}\boldsymbol{ }{{\varvec{M}}{\varvec{e}}{\varvec{t}}}_{{\varvec{k}}}+\boldsymbol{ }{\varvec{\varepsilon}}$$
(1)

The stepwise MLR estimated MDA8 O3 from meteorological factors (MDA8met); where y is the daily observed MDA8 O3 (MDA8obs) in season s (dry, wet, and all seasons) and city c (Jakarta and Bogor); Metk is the selected daily meteorological parameter; βs,c is the regression coefficient in season s and city c; βo is the intercept; and ɛ is the residual. MDA8obs used to construct the model were averaged from the MDA8obs at five sites in Jakarta and one for rural Bogor.

A series of steps was conducted to determine the relative meteorology contribution to O3 variation. First, correlation coefficients were established between MDA8obs and all meteorological parameters (see Supplementary Table S2). Second, the Variance Inflation Factor (VIF) was calculated to avoid multicollinearity between variables. Variables with a VIF > 10 indicated multicollinearity removal, and those with a 10 were retained. In this study, all the predictors had a VIF < 10 (approximately 1–3) and were retained for the next selection step. Third, stepwise regression was conducted to obtain the best model fit by adding or removing predictors based on Akaike Information Criterion (AIC) statistics. In this step, optimal meteorological parameters were selected and used to establish an MLR model. The optimal meteorological parameters, calculated intercept (βo), regression coefficient (βs,c), and adjusted R-squared for Jakarta (2013–2019) and Bogor (2017–2019) in the dry, wet, and all seasons are listed in Table 2, and Supplementary Table S3, S4. The MDA8 O3 resulting from the stepwise MLR model shows the contribution of the meteorological component (MDA8met). The difference between MDA8obs and MDA8met is called the residual and is considered an anthropogenic driver attributed to the ozone variability (MDA8ant)9,10,29.

The total change in meteorological contribution (ΔMDA8met) to O3 concentration can be calculated as follows:

$${{\varvec{\Delta}}\mathbf{M}\mathbf{D}\mathbf{A}8}_{\mathbf{m}\mathbf{e}\mathbf{t}}={\sum }_{{\varvec{k}}=1}^{{\varvec{n}}}{{\varvec{\beta}}}_{{\varvec{s}},{\varvec{c}}}\boldsymbol{ }{\varvec{x}}\boldsymbol{ }{{\varvec{\Delta}}\mathbf{M}\mathbf{e}\mathbf{t}}_{\mathbf{k}}$$
(2)

where ΔMetk is the change in the k-th meteorological variable. The change in MDA8ant (ΔMDA8ant) was the difference between the change in observed MDA8 (ΔMDA8obs) and ΔMDA8met. Thus, the relative contribution of meteorology to the ozone concentration could be calculated as the ratio of ΔMDA8met to the total meteorological and anthropogenic factors (ΔMDA8met + ΔMDA8ant).

Ozone formation sensitivity (OFS)

A common approach involves using satellite measurements to determine observed VOC-to-NOX ratios (FNR). FNRs have been widely used to study OFS. The tropospheric NO2 column reflects NOX emissions at the surface because most NO2 satellite measurements are within the planetary boundary layer and quickly transform or are removed from the atmosphere (short lifetime)14,18. HCHO is an intermediate product of all VOCs, and can be used as a proxy for VOC reactivity54. HCHO is a short-lived molecule that is usually found near emission sources.

The FNR threshold for separating the VOC-sensitive regime from the NOx-sensitive regime was derived using various methods for urban areas in the U.S. and China. The FNR threshold can vary by location owing to the influence of precursor emissions, meteorology, and geography17,55. This study compared FNR values in Jakarta and Bogor with those reported in the previous studies. The flowchart and methodology used in this study are shown in Supplementary Fig. S11.