Multiple Air Quality Monitoring Evidence of the Impacts of Large-scale Social Restrictions during the COVID-19 Pandemic in Jakarta, Indonesia

Air pollution is a top contributor to global mortality. Air quality issues abound in developing Asian countries, but during COVID-19 lockdowns, urban air quality improved due to the reduction in public mobility and fuel consumption. In Indonesia, the Large-Scale Social Restriction (LSSR) program was implemented to prevent the wider spread of COVID-19, especially in large urban areas. It was not a total lockdown program but had the purpose of reducing urban public mobility. This study investigated the effects of social restrictions on air quality in Jakarta, Indonesia. Data were obtained from our long-term monitoring of fine (PM2.5) and coarse particulate matter (PM2.5-10) and compositions collected at a site in South Jakarta. Other data were obtained from the environmental protection agency’s (EPA’s) air quality monitoring station in Central Jakarta including PM10, PM2.5, SO2, NO2, CO, and O3. The aerosol optical depth (AOD) in Jakarta measured by a sun photometer and satellite data were used to assess the spatial distribution of AOD across Jakarta. During the first LSSR implementation period from 15 March to 30 May 2020, there were decreased average SO2, CO, NO, NO2, and NOx concentrations of 40 to 60% compared to the same period in 2019. However, O3 increased by 33% likely due to reduction in NOx emissions. The PM2.5 decline reached ~40%, but a similar decline was not observed for PM10. Elemental and black carbon concentration data showed reductions that ranged from 30% to more than 50%. Consistent with the PM observations, both ground and satellite based AOD showed reductions in the aerosol column burden over the city. The ground based AOD values showed moderate correlations with PM2.5. The results confirmed that significant reduction in public mobility was highly associated with the improvement of local air quality which useful to derive future control strategies.


A C C E P T E D M A N U S C R I P T
4 burning fossil fuels, biomass, and the resulting emissions from vehicles and the power plants 69 (Brunekreef and Holgate, 2002). Breathing elevated concentrations of CO can reduce O2 transport 70 in hemoglobin and cause health effects including headaches, chest pain, heart disease, etc. (Sharma 71 et al., 1999) and moreover acute and chronic exposure in enclosed space will increase the risk for 72 development of cardiopulmonary and cardiovascular events, including death (Chen et al., 2007). 73 In Indonesia, air pollution is one of many serious environmental problems encountered major 74 cities due to the high population growth, increasing economic activity, intensive transport, and 75 industrial activities. 76 The capital city of Indonesia, Jakarta, is a megacity with a 2019 population of nearly 10.5 million 77 people. Jakarta is also surrounded by industrial and sub-urban areas that are located within the 78 range of 20-30 km from the center of the city. The city is characterized by a high mobility of 79 transportation that in turn routinely causes traffic jams. High urban PM2.5 pollution has been 80 observed with frequent violations of the Indonesian annual mean national ambient air quality 81 standard of 15 µg/m 3 (Santoso et al., 2013;Santoso et al., 2020). Twenty-four-hour PM2.5 values 82 at industrial sites ranged from 15 to 42 μg/m 3 , while at residential sites, values ranged from 9 to 36 83 μg/m 3 (Santoso et al., 2011). The PM2.5 concentrations measured in Jakarta at an arterial roadside 84 were higher than those measured at the other Indonesian big cities. The mean concentrations of

A C C E P T E D M A N U S C R I P T
5 contributed to the dangerous air quality index in Jakarta was surface ozone (O3) due to intensive 87 build-up of local precursor emissions (e.g. NOx) and the meteorological conditions favorable to 88 photochemical reactions (Permadi and Kim Oanh, 2008;Suhadi et al., 2005). In Jakarta, the mean 89 of NO2 concentrations using passive samplers showed the weekly average of 20-70 µg/m 3 , while 90 the peak daily concentration was 446 µg/m 3 . Annual average of NO2 in Asian cities typically in the 91 range of 23-74 µg/m 3 (WHO, 2005). Jakarta reported weekly average concentrations of SO2 92 between 4 μg/m 3 and 24 μg/m 3 (WHO, 2005). This value does not differ significantly from those 93 reported in central Jakarta in 2017. The reported mean SO2 was 22.72 µg/m 3 , while average NO2 94 was 11.85 µg/m 3 (Rahman and Barus, 2019). 95 In late 2019, a contagious virus appeared in China that was identified as a novel strain of 96 coronavirus belonging to the same family as acute respiratory syndrome (SARS) and Middle East 97 respiratory syndrome (MERS) (Liu et al., 2020;Zhu et al., 2020). Then, the so-called coronavirus 98 disease 2019  was declared by the World Health Organization (WHO) as a worldwide 99 pandemic in the mid-March 2020 based on the report of 118,319 cases and 4,292 deaths globally 100 (WHO, 2020). Although this novel virus is close to the coronavirus found in animals,  has been confirmed to be transmitted from human-to-human and has drawn significant attention 102

A C C E P T E D M A N U S C R I P T
9 particle concentration, density, refractive index and size. The sample filter is then placed on a white 160 standard and then measured its reflection value by repeating it three times (Lestiani et al., 2008;161 Diffusion system manufacture, 2012). The reflectance value obtained from the filter sample is a 162 value that is proportional to the number of BC in the filter using the assumption of an average of 163 particle mass absorption coefficient of 5.7 m 2 /g (Seneviratne, 2011). 164 The collected samples were analyzed for chemical composition using X-ray Fluorescence (XRF). 165 Multielement analyses were performed using an Epsilon 5 ED-XRF (Panalytical, Ltd) that has 9 166 secondary targets (Fe, CaF2, Ge, Zr, CeO2, Mo, Ag, Al and one Barkla polarizing target (Al2O3)). 167 Single element MicroMatter standards were used to develop the calibration parameters, while for 168 method validation, NIST SRM 2783 samples were periodically analyzed. These methods are 169 described in detail by Landsberger and Creatchman (1999) and Santoso and Lestiani (2014). XRF 170 analysis of the APM filters determined the concentrations of Na, Mg,Al,Si,S,K,Ca,Ti,V,Cr,171 Mn,Fe,Co,Ni,Cu,Zn,Pb and As. 172 The third location of the monitoring site is shown in Fig. 1 (C) and is surrounded by office 173 buildings and human settlements. Measurements of AOD using sun photometer were conducted by 174 the National Bureau of Meteorology, Climatology and Geophysics (BMKG) under the support of 175 The National Aeronautics and Space Administration (NASA), Aerosol Robotic Network

A C C E P T E D M A N U S C R I P T
10 1998). The measurement system is a solar-powered CIMEL Electronique 318A spectral radiometer 178 that measures sun and sky radiances at a number of fixed wavelengths within the visible and near 179 infra-red (VNIR) spectrum. 180 The stations are situated at a distance, for example, between station A and B are located of about 181 3 km while between station A and C there is a distance of about 4 km. The length span between 182 station C and B is more than 6 km therefore the selected stations represent different city background. 183

Statistical Analyses 184
To determine the changes in pollutant concentrations resulting from the reduced anthropogenic 185 activity during the COVID19 lockdown period, several non-parametric tests were performed to 186 compare the data from 4 periods: Pre-LSSR 2020 (January 1 to March 15, 2020), LSSR 2020 187 found to be not in a normal distribution and thus, comparisons among the 4 periods were done 190 using the Kruskal-Wallis ANOVA on ranks (Kruskal and Wallis, 1952). Individual pairwise 191 comparisons were made using the Bonferroni procedure (Bonferroni, 1936). To further confirm 192 these results, the data were also subjected to Mood's Median Test (Mood, 1954

A C C E P T E D M A N U S C R I P T
12 215

Particulate matter 216
The distributions of PM10 and PM2.5 derived from hourly data for all 4 periods are presented as 217 box and whisker plots in Fig. 2. The average values of PM10 for Pre-LSSR 2020 and LSSR 2020 218 were 41.0±18.4 and 49.4 ± 20.3 µg/m 3 , respectively. The average values of PM2.5 for Pre-LSSR 219 2020 and LSSR 2020 were 23.8 ±16.7 and 26.5 ± 15.5 µg/m 3 , respectively. PM10 and PM2.5 were 220 higher in the LSSR 2020 period than in the Pre-LSSR 2020 period. This difference may be the 221 result of much higher total rainfall in the Pre-LSSR 2020 period (836 mm) than any of the other 3 222 periods (195,290, and 207 mm, respectively, for the other 3 periods). Period average values of 223 rainfall intensity (in mm) for different periods is presented in Fig. S1. Precipitation reduces PM2.5 224 concentrations at a lower extent than it reduces PM10 concentrations (Blanco-Becerra et al, 2015;225 Zhou et al., 2020). Comparing the LSSR period in 2019 with the LSSR 2020 period showed that 226 the average PM2.5 values were 45.4 ± 21.3 and 26.5 ± 15.5 µg/m 3 , respectively, while PM10 227 concentrations were 46.1 ± 18.8 and 49.4 ± 27.3 µg/m 3 , respectively. PM2.5 in 2019 was 228 substantially higher than in 2020. However, PM 10 in 2019 was slightly lower than in 2020. This 229 difference could be due to the somewhat higher rainfall intensity in 2019 (Fig. S1), while Showing 230 that the LSSR in 2020 had a greater effect on PM2.5 than on PM10. Thus, there are other sources of 231

13
To examine these temporal patterns in more detail, the hour-by-hour data have also been 233 analyzed for differences over these 4 defined periods. Fig. 3 show the hour-by-hour box and 234 whisker plots showing the distributions of PM2.5. The maximum diel average of PM2.5 also showed 235 consistent result with the mean values with lower value during the LSSR period in 2020 (LSSR 236 2020) as compared to both periods in 2019. However, the Pre-LSSR 2020 value was lower than 237 the LSSR 2020 period due to high rainfall (Fig. S1). For PM10, the maximum diel average value of 238 LSSR 2020 period was higher as compared to all other periods (Fig. S2). Typical diel pattern of 239 PM10 was seen for all periods with higher values during daytime. For PM2.5, higher values were 240 seen at the late night may reflect the lower dispersion characteristics overnight when wind speeds 241 and mixed layer heights would lead to lower dilution of ground level emissions. The higher daytime 242 PM10 concentrations may reflect higher windspeeds to suspend coarse mode particles beginning 243 after sunrise. The concentrations decreased in the afternoon as the mixed layer heights increase. 244 245

Gases (SO2, CO, NO, NO2, NOX, and O3) 246
Box and whisker plots for the gaseous pollutants are given in Fig. 4. SO 2 , CO, and the oxides of 247 nitrogen (NO, NO2, and NOX) were substantially lower during the LSSR 2020 period compared to 248 the prior months in 2020. The LSSR 2019 period had higher CO, SO2, NO, NO2, and NOX than in 249

A C C E P T E D M A N U S C R I P T
14 experienced a significant decrease since people was encouraged to stay at home due to the virus. 252 Many sources reported that declines in transportation by rail (7%), sea public transportation 253 (50.7%), air transportation (82.4%) and private car (19.3%). The mobility of the people going to 254 the market and pharmacy fell by 67%. Those going to the mall/café were down by 77%. The 255 decrease in travel wass expected to be higher in parallel with the increases in positive cases of 256 COVID-19 (Caraka et al., 2020). Thus, the emissions of SO2, CO, and NO2 were also reduced. NO2 257 showed its lowest value during the LSSR 2020 period. This difference suggests that these species concentrations were highest during the LSSR 2020 period. Meteorological factors also affected the 267 measured ozone concentrations as well as the VOC emissions from local and regional non-traffic 268

A C C E P T E D M A N U S C R I P T
15 (e.g., Mahato et al., 2020;Chen et al., 2020;Abdullah et al., 2020;Qiu et al., 2021;Huang et al., 270 2021). The increased ozone suggests that Jakarta is in a VOC-limited regime for ozone formation 271 (Seinfeld and Pandis, 2016) so that reducing NO emissions led to increased ozone. There are 272 relatively small differences in the monthly average photoperiods ranging from 12.4 h in January to 273 11.8 h in June so that variations in photochemical activity and temperature are much less than in 274 many other locations observing increased ozone concentrations. 275 We also compared the diel patterns between LSSR 2020 period and corresponding period in hourly distributions. The CO morning rush hour peak was distinctly lower during LSSR 2020 than 290 any of the other 3 periods suggesting that the reduced mobility period was successful in reducing 291 the light duty traffic volume. The diel pattern of SO2 distributions in the Pre-LSSR periods showed 292 relatively uniform hourly values over the whole day. In both LSSR periods, there were small 293 increases from 8:00 to noon and a drop into the afternoon. Diesel fuel in Indonesia has a high sulfur 294 content (3000 ppm) so they represent a local SO2 source. However, heavy-duty diesel trucks are 295 restricted to overnight hours. Thus, this daytime rise seems likely to be the result of downmixing 296 of emissions from the stacks of coal-fired power plants and oil refineries in West Java (Santoso et 297 al., 2020). These results strongly suggest that the restricted mobility rules reduced motor vehicle 298 traffic and some industrial activities resulting in decreased concentrations of CO, SO2, and NOX. 299 However, there were increasing O3 concentration that can be understood in terms of decreased NO 300 titration and sufficient reactive hydrocarbon concentrations to support ozone formation. 301 302 3.2 PM2.5-10 and PM2.5 mass and multi-elemental compositions 303 This section presents the results of our long-term PM and composition monitoring to investigate 304 the impact of LSSR on PM air quality in Jakarta (Santoso et al., 2020). Both coarse and fine PM 305 monitoring results are presented and discussed separately.

A C C E P T E D M A N U S C R I P T
17 307

Coarse PM (PM2.5-10) 308
We compared the period average (derived from daily average concentrations) of PM2.5-10 and 309 compositions between pre-LSSR and LSSR period and the results are presented in Table 3  for the same months. The results are presented in Fig.7a. The monthly averages of PM2.5-10 during

A C C E P T E D M A N U S C R I P T
18 was only 10 µg/m 3 while the long-term average value was 27 µg/m 3 . Similar results were seen for 326 April and May where the long-term average concentrations were more than double of the 2020 327 measurements confirming the effect of the LSSR on coarse PM in Jakarta. 328 329

Fine PM (PM2.5) 330
Period averages of PM2.5 and elements were compared for pre-LSSR and LSSR periods and the 331 results are presented in Table 4  LSSR 2020 decreased by more than 50% from 1333 ng/m 3 to 543 ng/m 3 , but the concentration did 338 not change compared to pre-LSSR 2020. Average S concentration was similar between the LSSR 339 2020 period and the pre-2020. The value was also lower than in the LSSR 2019 period. BC 340 concentrations declined by more than 30% due to the reduction of on-road transport. Compared to 341 the 2019 period, there was a decrease in the concentration of the elements Pb, Zn, Cu, Fe in the 342

A C C E P T E D M A N U S C R I P T
19 concentrations during the LSSR 2020 period was a result of the reduced anthropogenic activities 344 during the LSSR implementation in Jakarta. 345 Long term monthly average PM2.5 concentrations were calculated for March, April and May 346 during the period of 2010-2019. We then compared the values with those calculated for the year of 347 2020 when the LSSR was implemented and the result is presented in Fig. 7b. Typical reduction of 348 concentration during the LSSR was also clearly seen. In all months, the PM2. affected by the number of daily data available hence data completeness is important. The AOD 360 data were correlated with the ground PM measurements at the same site. The data provided an

A C C E P T E D M A N U S C R I P T
20 indication that the improvement of PM air quality during the LSSR period was also captured by 362 the ground based AOD observations. 363 The AOD was compared with the PM2.5 concentrations obtained from the two sites presented 364 above (filter-based (daily) and continuous monitoring) for the period of Oct 2019 -May 2020 (See 365 Fig. S4). Comparison of more than 25 data pairs for the second site showed relatively good 366 correlation between AOD and our PM2.5 data (showed by coefficient of determination value, r 2 of 367 0.596). A comparison for more than 100 data pairs between AOD and PM2.5 measured at EPA's 368 automatic monitoring station also showed a moderate correlation with r 2 = 0.453. Therefore, there 369 was a consistent PM reduction as shown by these data. These findings supported the results from the ground AOD observations. Larger scale 377 observation by satellite observed similar AOD reduction patterns especially above the Jakarta area.

A C C E P T E D M A N U S C R I P T
21 fire hotspot was seen over the provinces of Riau and Jambi (Sumatera Island). However, it would 380 not affect Jakarta's air quality due to the southwest monsoon synoptic winds. The reduction was 381 also seen for other areas especially in the western part of Java island. 382 383

385
Various air quality observations combining ground-based and satellite derived data were utilized 386 to investigate the impact of the LSSR 2020 on urban air quality in Jakarta, Indonesia. Continuous 387 monitoring data from the AQMS installed in Central Jakarta showed reductions in PM2.5, NO2, 388 CO, and SO2 concentrations during the LSSR period as compared to the period before (normal). 389 However, ozone increased in the LSSR 2020 period. Our long-term PM monitoring at another site 390 located in the southern part of the city showed consistent substantial reductions of coarse and fine 391 PM as well as major elements during the LSSR period. The findings were enhanced by the ground-392 based and MODIS Terra AOD observations which showed exceptional lower AOD values during 393 the period where traffic and other anthropogenic activities were reduced. While the results 394 indicated that the government's program was rather successful as seen from the air quality 395 monitoring data, the situation maybe different in other areas where the data do not exist. It is 396 suggested that this study should be followed up by the epidemiological research to investigate the 397 potential short-term health benefits. 398