Beehive Fireworks Festival Effect on the Nearby Atmospheric PM 2.5 Level

Sudden short-term air pollution episode is now widely considered to harm human health. Previous research has found that firework activities rapidly raise the PM 2.5 level in the ambient air. This study investigates the influence of Yanshuei Beehive Fireworks Festival on atmospheric PM 2.5 from February 9 th to 12 th , 2017. The PM 2.5 samples were gathered at 8 sampling sites around Yanshuei and Xinying before (background (B)), trial (T), during (D), and after (A) beehive firework display periods during the Yanshuei Beehive Fireworks Festival. The temporospatial differences of atmospheric PM 2.5 before and after fireworks activities were explored. The atmospheric PM 2.5 level in major activity areas was significantly higher from the background level in the trial and festival periods, and even after the activity. The study revealed that PM 2.5 level reached the highest value of 327 µ g m –3 at the major activity areas, which is 6.6 and 5.9 times those at upwind (49.8 µ g m –3 ) and downwind (55.5 µ g m –3 ) sites, respectively. Additionally, the T/B and D/B ratios were 3.01 and 7.19, respectively, around the major activity area. Conversely, the wind rose diagrams and contour lines of PM 2.5 concentrations evaluated using Surfer 10.0 around the ambient air demonstrate that the atmospheric PM 2.5 levels at Yanshuei and Xinying were similar to each other (35–45 µ g m –3 ). However, the PM 2.5 hardly diffused to distant places and accumulated in the local area around the boundary between Yanshuei and Xinying during the Beehive Fireworks Festival, since the wind speed was usually low or even stayed calm. The iso-concentration contour maps show that K + , Cl – , Mg 2+ , and NO 3– are related to firework release during the festival.


INTRODUCTION
The health impacts of short-term air pollution (including fireworks, temple fair, vehicle exhaust at traffic intersections, welding fumes, incense smoke, and even fire smoke) have garnered significant attention in recent years, as evidenced by several investigations characterizing anthropogenic emissions, especially in densely populated urban regions (Chang et al., 2013;Kulshrestha et al., 2004;Lin et al., 2016;Ravindra et al., 2003;You et al., 2013;Tseng et al., 2022;Hien et al., 2022;Collado et al., 2023;Debnath et al., 2022;Wang et al., 2022). The widespread application of pyrotechnics at massive celebratory events often deteriorates short-term air quality, potentially endangering public health (Gorbunov et al., 2013;Smith and Dinh, 1975;Pang et al., 2021;Yu et al., 2021). Globally, the firework display has evolved as a source of atypical atmospheric pollution. A study on fireworks displays indicated that about 80% of the particulate matter emitted by the explosive content of the fireworks is in the PM2.5 fraction (Keller and Schragen, 2021).
The physicochemical characteristics of fireworks have been studied extensively in numerous studies. Many investigations have observed that high concentrations of gaseous pollutants can be observed during firework periods (Ambade, 2018;Zhang et al., 2017). Some studies have investigated the formation of secondary inorganic ions during Spring Festival firework displays (Jiang et al., 2015;Wang et al., 2019). The fine particulate nitrate formed during the firework display is mainly due to the heterogeneous reaction of HNO3 with NH3, which is similar to the nitrate formation mechanism during haze period (Tian et al., 2019;Wu et al., 2018).
Numerous works have documented elevated concentrations of water-soluble ions (K + and Cl -) in ambient aerosol particles during and shortly after fireworks displays (Drewnick et al., 2006;Godri et al., 2010;Moreno et al., 2007;Ravindra et al., 2003;Shen et al., 2009;Tsai et al., 2012;Wang et al., 2019;Yang et al., 2014;You et al., 2013). However, the impact of short-term air pollution of PM2.5 and chemical compositions (water-soluble ions components/concentrations) from beehive firework displays in the Yanshuei area of southern Taiwan have seldom been explored. Therefore, this study investigates the PM2.5 concentrations, and determines the watersoluble ion components (Na + , K + , NH4 + , Mg 2+ , Ca 2+ , Cl -, NO3 -, and SO4 2-) and concentrations from the samples gathered during Beehive firework displays in the Yanshuei area of southern Taiwan. The mass concentration and water-soluble ions for PM2.5 were determined from the samples accumulated by eight PQ200 samplers. This investigation adopted the pollution model (i.e., inverse distance to a power method) to analyze the water-soluble ions and PM2.5 around the Yanshuei area. Analytical results indicate that the impact range of PM2.5 from Beehive fireworks is approximately 4-6 km, while that of water-soluble ions is only 2 km. The beehive fireworks display degrades short-term air quality, thereby requiring attention to address health concerns.

Collection of Particulates
Atmospheric particulate samples were collected in the Yanshuei area of southern Taiwan during the Lantern Festival from February 9 to 12 in 2017. The PM2.5 samplings were carried by PQ200 at 8 sampling sites, encompassing the concentrated fire area plus windward and leeward sites. The sampling sites of the concentrated fire area were the Wumiao Temple (Z1) and Yanshuei police office (Z2). The Wumiao Temple (Z1) sampling site was located on the roof of a three-story building (9 m height), roughly 50 m north of the major beehive fireworks display site, while the Yanshuei police office sampling site was located on the roof of a four-story building (12 m height), roughly 300 m south of the major beehive fireworks display site. The windward sampling sites were in the Nanrong University of Science and Technology (W1), Kwanai Elementary School (W2) and Xuesi Garden (W3), each located on the roof of a building (12 m, 8 m and ground height, respectively), roughly 850 m east, 850 m northwest and 550 m northwest of the major beehive fireworks display site, respectively. The leeward sampling sites were in the Mingda high school (L1), Prince Junior High (L2) and Renguang Elementary School (L3), each located on the roof of a building (12 m, 4 m and ground height, respectively), roughly 1,700 m southeast, 3,650 m south and 2,700 m south of the major beehive fireworks display site, respectively. Fig. 1 shows the locations of the sampling. The Yanshuei beehive fireworks display events occurred within the four stages of our experimental periods. This investigation considered February 9 th (12:00)-10 th (12:00), 10 th (12:00)-11 th (18:00), 11 th (18:00-24:00) and 12 th (0:00-12:00), 2017 as the Before (background), Trial, During and After beehive firework display periods, respectively. The average air temperature, relative humidity and wind speed during the sampling period were 13.8°C (10.2°C-20.9°C), 65.5% (49.0%-90.0%), and 0.25 m s -1 (0.0 m s -1 -3.6 m s -1 ), respectively (with no rain). PM2.5 samples were accumulated using eight low-volume ambient air samplers (PQ200 by BGI, Inc., USA), equipped with Teflon (diameters, 47 mm). The PQ200 sampling flow rate for PM2.5 was 16.7 L per min. Before and after each sampling, filters were dried for 24 hours in a desiccator at 25°C ± 3°C and 40% ± 5% relative humidity. They were then weighed on an electronic balance (UXM2; Mettler Toledo) with a precision of 0.1 µg. Finally, the concentration of suspended particulate material was calculated by dividing the particle mass by the volume of tested air. Table 1 shows the weather data during the sampling period. These data were accumulated using Davis Vantage Pro 2 Weather Stations MODEL 6152C professional wired weather meter. The Davis Vantage Pro 2 weather monitoring system comprises an integrated weather sensor, which can record complete and real-time meteorological data (including time, wind direction, wind speed, temperature, relative humidity and atmospheric pressure) on the sampling periods. The wind speeds during the Background (B), Trial (T), During (D) and After (A) sampling periods were 0.0-3.6 m s -1 (0.33 ± 0.61 m s -1 ), 0.0-0.9 m s -1 (0.07 ± 0.20 m s -1 ), 0.0-0.4 m s -1 (0.07 ± 0.15 m s -1 ), and 0.0-1.3 m s -1 (0.18 ± 0.31 m s -1 ), respectively. The wind direction in each period was NNE, NNE, NW and NW, respectively; the static wind rate was 56%, 82%, 83% and 68%, respectively.

Water-soluble Ion Analysis and Quality Control
To measure the levels of particle-bound water-soluble inorganic species (Na + , K + , NH4 + , Mg 2+ , Ca 2+ , Cl -, NO3 -, and SO4 2-) were determined, the PM2.5 samples were extracted from each Teflon fiber filter with 10 mL of ultra-pure water (specific resistance 18.3 MΩcm). The water-soluble ions were extracted using an ultrasonic bath (UC-300) for 120 minutes. All extraction solutions were filtered through a cellulose acetate filter (ADVANTEC MFS, Inc., USA cat No., CO20A025A; pore size 0.2 µm; diameter 25 mm) and stored in plastic vials in a refrigerator at 4°C until they were chemically analyzed. The inorganic species were analyzed by ion chromatography (IC) (DIONEX ICS-3000 with conductivity detection (DC detector/chromatography module, P/N 061767)).
The cations were quantified using a DIONEX IonPac ® 4 × 50 mm CG12A guard column, a DIONEX IonPac ® 4 × 250 mm CS12A analytical column and a cation self-regenerating suppressor (CSRS ® ULTRA II, 4 mm, AutoSuppression ® Recycle Mode); The anions were quantified by a DIONEX IonPac ® 4 × 50 mm AG11 guard column, a DIONEX IonPac ® 4 × 250 mm AS11 analytical column and an anion self-regenerating suppressor (ASRS® ULTRA II, 4 mm, AutoSuppression® Recycle Mode). The eluents for the cation and anion analyses were mixed with 20 mM of methane sulfonic acid and 12 mM of NaOH, respectively. Analytical drift was monitored throughout the analytical procedures. Recovery efficiencies were then determined using diluted samples that were spiked with known quantities of the ions of interest. Recovery efficiencies ranged from 93% to 107%. The method detection limit (MDL) was estimated by repeatedly analyzing a control solution of known quality. Replicate analysis of IC measurements was performed to calculate the MDL of each element using MDL = 2.681 × Spooled, with SA 2 / SB 2 < 3.05. Spooled = [(6SA 2 + 6SB 2 )/12] 0.5 , where Spooled denotes the pooled standard deviation; SA represents the standard deviation of the prepared sample with a larger F-test value, and SB is the standard deviation of the other sample. The detection limits were expressed in ng m -3 (estimated from MDL × volume of analyzed solution (10 mL)/average sampling volume (20 m 3 )) were Na + , 7.00 ng m -3 ; K + , 12.0 ng m -3 ; NH4 + , 27.0 ng m -3 ; Mg 2+ , 13.0 ng m -3 ; Ca 2+ , 10.5 ng m -3 ; Cl -, 46.6 ng m -3 ; NO3 -, 28.6 ng m -3 , and SO4 2-, 61.6 ng m -3 . Both field and laboratory blank samples were prepared and analyzed for each sampling and analysis. All data were corrected with blanks.

Data Processing
The wind rose analysis: Microsoft Excel was adopted to place wind rose figures in the sampling sites of concentrated fire areas based on data of wind direction, wind speed and frequency.
Trend analysis of PM and water-soluble ion: Surfer ® 2D&3D mapping, modeling and analysis software (Version 10) was adopted to draw iso-concentration contour maps showing the trend of PM and water-soluble ions in the Yanshuei area. This investigation adopted the gridding method of inverse distance to power to paint an iso-concentration contour map by gathering all data of PM and water-soluble ions.

RESULTS AND DISCUSSIONS
3.1 Concentrations of PM2.5 Table 2 shows the concentration of PM2.5 around the Yanshui area (i.e., leeward sites (L), windward sites (W) and central zone (Z)) at 4 events (i.e., Background (B), Trial (T), During (D) and,  New Year's Eve, New Year and Lantern Festival. The concentration of PM2.5 on the leeward site (L2) was significantly higher than on windward sites (W1-W3). Lin et al. (2014) used MOUDI to monitor concentrations of PM2.5 in Z1 (windward) and Z2 (leeward) sites which were 165 µg m -3 and 437 µg m -3 , respectively. The graphs in Fig. 2 indicate that PM2.5 concentration was higher in period T than in D in the windward and leeward sites, except the W3 site (near the central zone). The reason may be to avoid disturbance from major fireworks in the central zone during the firework trial. In addition, the concentration of PM2.5 in the central zone was highest during period D. The concentrations of PM2,5 from the EPA monitoring stations (i.e., SY, AN, SH and TN) were also compared. The concentration of PM2,5 was sequentially T > D > A > B periods in the SY and SH stations. The concentration in the AN and TN stations was in the order D > T > A > B periods sequentially. The reason for the difference in sequence is that the SY and SH stations are closer to the Yanshuei area, resulting in a higher concentration of PM2.5 during period T than during period D.

Trial/Background (T/B), During/Background (D/B) and After/Background (A/B) Ratios of PM2.5 Concentration
To investigate the concentration trend of PM2.5, results from the background period were compared with the results from the trial (i.e., T/B ratio), during (i.e., D/B ratio) and after (i.e., A/B ratio) periods, as presented in Table 3. The T/B and D/B ratio ranges at windward sites were 1.22-1.33 and 1.11-1.26, respectively. The T/B and D/B ratios from site Z2 were 3.01 and 7.19, respectively. The T/B and D/B ratio ranges at the leeward sites were 1.49-1.74 and 1.41-1.44, respectively. The T/B and D/B ratios were significantly higher at Z2 than at the windward and leeward sites, and the leeward sites had higher ratios than the windward sites. Lin et al. (2014) found that the T/B and D/B ratios during the Yanshuei beehive firework display in 2013 were 4.3 and 15.5, respectively at site Z2, and 1.8 and 6.3 at site Z1, and thus were higher than our results (1.53 and 2.05, respectively). The amount of garbage generated by fireworks during the Yanshuei beehive firework performance increased by a factor of 3.21 between 2013 and 2016, from 29.4 tons to 94.5 tons. These findings suggest that fewer fireworks were used during the firework performance in 2013 than in 2016, resulting in reduced PM2.5 concentrations. Pervez et al. (2016) monitored the impact of firework displays on air quality during the Diwali festival in India, and found that concentration of PM2.5 reached 1,501.20 µg m -3 , 831.98 µg m -3 and 397.12 µg m -3 from November 13 to 15, respectively, which was 2 to 8 times that on typical days (i.e., without any firework display).
From Table 3, A/B < 1 at all sites except Z2 and L2. The T/B and D/B obtained from the EPA monitor station ranged from 1.35 to 1.65 (mean 1.48), revealing that the beehive firework display significantly affected the air quality around the Yanshui area, increasing the concentration of PM2.5 by about 48%. Fig. 3 shows the wind roses and iso-concentration contour maps combining the wind direction with. In period B, the PM2.5 concentrations ranged from 35 µg m -3 to 45 µg m -3 , and the wind rose had six areas of higher concentration, all with similar concentration levels. The PM2.5 concentration in period T ranged from 40 µg m -3 to 130 µg m -3 , and the highest concentration was recorded in the Z2 site. The concentration gradient during period T was about −45 µg m -3 km -1 and the impact distance was about 2 km. In the D periods, the PM2.5 concentration ranged from 60 µg m -3 Table 3. T/B, D/B, and A/B values of PM2.5 concentration at each sampling site during the sampling period.

Item
The ratios of PM2 to 320 µg m -3 . The concentration gradient during the D period was about −65 µg m -3 km -1 and the impact distance was 4 km. The concentrations of PM2.5 in period A ranged from 50 µg m -3 to 145 µg m -3 . The concentration gradient during the period A was about −32 µg m -3 km -1 , and the impact distance was around 3 km. Owing to the high static wind ratio and slow wind speed during the beehive firework display, the PM2.5 particles were not easy to diffuse. Therefore, the beehive fireworks mainly affected the Yanshuei area and the junction within the Xinying district.

The Concentration of Water-Soluble Ions in PM2.5 during the Yanshui Beehive Fireworks
Figs. 4-7 show the concentration of K + , Cl -, Mg 2+ , and NO3 -, which indicated that the concentration distributions of Cl -, K + , and Mg 2+ were similar during the 4 periods and the highest concentration of water-soluble ions was observed in site Z2. The concentration of K + in PM2.5 during the 4 periods (B, T, D, A) were sequentially 2.29 µg m -3 , 23.8 µg m -3 , 99.1 µg m -3 , and 33.4 µg m -3 in the Z2 site. These results surpass those of other sites (without Z2) by a factor of 4.32, 13.2, 39.7 and 25.0, respectively. In the Z2 site, the concentration of Clin PM2.5 during the 4 periods were sequentially 1.01 µg m -3 , 14.1 µg m -3 , 46.1 µg m -3 , and 16.8 µg m -3 which is 1. 47, 8.20, 16.2, and 13.6 times those of other sites, respectively. Additionally, the concentration of Mg 2+ in PM2.5 during the 4 periods were sequentially 0.0526 µg m -3 , 0.0996 µg m -3 , 0.2998 µg m -3 , and 0.1252 µg m -3 which is 0.91, 12, 10 and 5.9 times those in other sites, respectively. Wilkin et al. (2007) and Shi et al. (2011) observed a large amount of Clin the particles after a firework display in the urban area. Li et al. (2013) and Wu et al. (2018) pointed a larger amount of K + in the particles corresponds to a large amount of Cl -, because the particles from fireworks contain KCl, KClO4 and chlorine organics. KNO3 was used as a propellant, while Al and Mg were commonly used as fuels and white luminous agent in fireworks (Lorenzo et al., 2021). Some research from Beijing (Cheng et al., 2014), Taiwan (Tsai et al., 2012) and New Delhi (Kumar et al., 2016) indicated that the concentration of Cland K + increased rapidly after a firework display, since compounds of potassium with nitrate, chlorate and perchlorate are widely adopted as oxidizers in fireworks. The T/B, D/B and A/B of Cl -, K + , and Mg 2+ in PM2.5 from the Z2 site were sequentially 14, 46, and 18 for Cl -; 10, 43 and 15 for K + , and 25, 64 and 25 for Mg 2+ . Fig. 7 shows the iso-concentration contour map of NO3in PM2.5, indicating that NO3concentration was higher in periods T and D than in B and A. The concentration of NO3was obviously 7.9 times higher than other sites (mean 8.61 µg m -3 ) than in Z2 (67.7 µg m -3 ). The T/B, D/B, and A/B of  NO3in PM2.5 from the Z2 site were 1.8, 4.9 and 1.5 respectively. This may be attributed to the large amount of beehive fireworks the Z2 site, and the fireworks may contain a small amount of Ba(NO3)2. The concentration of NO3in Z2 was 1.1 and 3.1 times that than other sites during periods T and A. In the other sites, the concentration of NO3was lower in periods D and A than in period B. The reason may be that the proportion of static wind was relatively high, and pollutants did not diffuse easily, thus leading secondary pollutants to be generated in situ. The concentration of SO4 2-, NH4 + , Na + , and Ca 2+ did not increase significantly in the Z2 site (Figs. S1-S4; Figure numbers preceded by an "S" are in the Supplementary Materials).

The Iso-concentration Contour Map of Water-Soluble Ions in PM2.5 during the Yanshui Beehive Fireworks
To study the influence of the Yanshui beehive firework display on air quality around the Yanshui area, the concentration of water-soluble ions in PM2.5 was monitored. Figs. 8-11 show the iso-concentration contour maps of K + , Cl -, Mg 2+ , and NO3 -. These results indicated that isoconcentration contour maps of K + , Cl -, Mg 2+ , and NO3during periods T, D, and A were different from those in period B, and the concentration gradually reduced from Z2 to the periphery, which was more consistent with the iso-concentration contour map of PM2.5. However, the range of the iso-concentration contour line from Z2 was about 2 km, which is smaller than the 3-4 km range of PM2.5. The possible reason is that the average wind speed was around 0.07 m s -1 during period D, and the static wind ratio was as high as 83%. The resulting indicator pollutant accumulated in Z2 and formed a secondary pollutant; its range of influence was only 2 km. The SO4 2and NH4 + ions in the secondary aerosol behaved similarly, but do not have significantly higher concentrations than other iso-concentration groups (Figs. S7-S8; Figure numbers preceded by an "S" are in the Supplementary Materials). This may be attributed to the reason that thermal NOx emitted by the beehive fireworks display is converted into NO3at a faster rate than SO4 2and NH4 + , and it is mainly distributed on ultrafine particles Lin et al., 2009). The iso-concentration group of NO3in Z2 were higher than in surrounding areas during periods T, D and A, while iso-concentration group of SO4 2and NH4 + showed the opposite trend (i.e., concentration was higher in the surrounding than in the central zone).

CONCLUSION
The concentrations of fine particles were significant higher during the Yanshuei Beehive Firework Festival than in non-festival times. The PM2.5 level reached the highest value in the central area of the firework display (Z2) of 327 µg m -3 , which was 6.6 and 5.9 times than that of levels at upwind (49.8 µg m -3 ) and downwind (55.5 µg m -3 ) sites, respectively. Additionally, the T/B and D/B ratios were 3.01 and 7.19, respectively, around the major activity area. The concentration distributions of Cl -, K + , and Mg 2+ were similar during the 4 periods and the highest concentration of water-soluble ion was observed in the Z2 site. The secondary aerosol concentration of NO3in Z2 (67.7 µg m -3 ) was 7.9 times that in other sites (mean 8.61 µg m -3 ). The iso-concentration contour maps also show that the beehive firework festival produced K + , Cl -, Mg 2+ , and NO3 -.