Environ Sci Pollut Res DOI 10.1007/s11356-016-7177-0
RESEARCH ARTICLE
Gas-phase ammonia and water-soluble ions in particulate matter analysis in an urban vehicular tunnel Marcelo S. Vieira-Filho 1 & Debora T. Ito 2 & Jairo J. Pedrotti 2 & Lúcia H. G. Coelho 3 & Adalgiza Fornaro 1
Received: 15 October 2015 / Accepted: 4 July 2016 # Springer-Verlag Berlin Heidelberg 2016
Abstract Ammonia is a key alkaline species, playing an important role by neutralizing atmospheric acidity and inorganic secondary aerosol production. On the other hand, the NH3/NH4+ increases the acidity and eutrophication in natural ecosystems, being NH3 classified as toxic atmospheric pollutant. The present study aims to give a better comprehension of the nitrogen content species distribution in fine and coarse particulate matter (PM2.5 and PM2.5–10) and to quantify ammonia vehicular emissions from an urban vehicular tunnel experiment in the metropolitan area of São Paulo (MASP). MASP is the largest
Responsible editor: Gerhard Lammel Electronic supplementary material The online version of this article (doi:10.1007/s11356-016-7177-0) contains supplementary material, which is available to authorized users. * Adalgiza Fornaro
[email protected] Marcelo S. Vieira-Filho
[email protected] Jairo J. Pedrotti
[email protected] Lúcia H. G. Coelho
[email protected] 1
2
3
Departamento de Ciências Atmosféricas, Universidade de São Paulo, Rua do Matão, 1226, Cidade, Universitária, São Paulo, SP 05508-090, Brazil Escola de Engenharia, Universidade Presbiteriana Mackenzie, Rua Consolação, 896, Consolação, São Paulo, SP 01302-907, Brazil Centro de Engenharia Modelagem e Ciências Sociais Aplicadas, Universidade Federal do ABC, Avenida dos Estados, 5001, Bangu-Santo André, SP 09210-170, Brazil
megacity in South America, with over 20 million inhabitants spread over 2000 km2 of urbanized area, which faces serious environmental problems. The PM2.5 and PM2.5–10 median mass concentrations were 44.5 and 66.6 μg m−3, respectively, during weekdays. In the PM 2.5 , sulfate showed the highest concentration, 3.27 ± 1.76 μg m−3, followed by ammonium, 1.14 ± 0.71 μg m−3, and nitrate, 0.80 ± 0.52 μg m−3. Likewise, the dominance (30 % of total PM2.5) of solid species, mainly the ammonium salts, NH4HSO4, (NH4)2SO4, and NH4NO3, resulted from simulation of inorganic species. The ISORROPIA simulation was relevant to show the importance of environment conditions for the ammonium phase distribution (solid/aqueous), which was solely aqueous at outside and almost entirely solid at inside tunnel. Regarding gaseous ammonia concentrations, the value measured inside the tunnel (46.5 ± 17.5 μg m−3) was 3-fold higher than that outside (15.2 ± 11.3 μg m−3). The NH3 vehicular emission factor (EF) estimated by carbon balance for urban tunnel was 44 ± 22 mg km−1. From this EF value and considering the MASP traffic characteristics, it was possible to estimate more than 7 Gg NH3 year−1 emissions that along with NOx are likely to cause rather serious problems to natural ecosystems in the region. Keywords Ammonia vehicular emissions . Fine and coarse particulate matter . Megacity . ISORROPIA
Introduction One of the main challenges regarding the air pollution evaluation in modern metropolis consists of narrowing uncertainties concerning the gases and particulate matter vehicular emissions (Gillies et al. 2001; Sánchez-Ccoyllo et al. 2009;
Environ Sci Pollut Res
Suarez-Bertoa et al. 2015). Chassis dynamometer or engine emission measurement studies are important to quantify the exhaust emission from vehicles. However, these studies are limited to specific vehicles and can neither simulate the diversity of sources in an urban atmosphere nor estimate nonexhaust particles arising from abrasive sources (Lawrence et al. 2013). Several issues are important when dynamometer studies are taken into consideration, e.g., vehicle wear (mileage) and catalyst technology. The tunnel studies represent a realistic snapshot of the vehicular fleet condition in a given moment (Mancilla and Mendoza 2012), providing detailed characteristics such as the total number and units per hour which could be extrapolated to better understand how the average traffic logistics behaves in a specific environment at a specific time (Pérez-Martínez et al. 2014). It is important to consider that vehicular tunnel studies are also a keen tool to improve pollutant inventories with substantial field data (Colberg et al. 2005; Staehelin et al. 1997; Zhou et al. 2014). Ammonia (NH3) is an atmospheric trace alkaline gas that plays an important role in the atmospheric acidity neutralization, generating ammonium (NH4+) salts which are the main secondary inorganic aerosol source (Bouwman et al. 2010; Ianniello et al. 2010; Zhang et al. 2015). Ammonium salts, considered 10-day lifetime stable aerosols, are classified as fine matter particulate, PM2.5, with diameter 10 (not representative). It is important to note that the AIC method does not evaluate null hypothesis or its statistical significance for a single model, but the AIC indicates the fittest model in a series for the same dataset. For the scope of this study, the stepwise analysis was applied for the water-soluble species of the fine particulate matter dataset. In addition, this robust model predicts fine particulate matter ammonium concentration, describing the ionic species which have a major role in particulate matter for both sites PP and JQ.
HNO3 to estimate water content, aerosol salt concentrations, and gaseous aerosol precursors. Despite the forward or reverse problem, the system aerosol-atmosphere could be considered either thermodynamically stable (in which precipitation of salts occurs) or metastable (in which the system is considered a supersaturated aqueous solution). For the scope of this study, all the simulations performed were considered thermodynamically stable in order to solve a reverse problem. In addition, the model was applied to evaluate the distribution of nitrogen species (ammonium salts) in the urban fine particulate matter.
Results and discussion Throughout the sampling campaign period, the average temperature and relative humidity registered outside the tunnel (PP) were 18.6 ± 2.7 °C and 82 ± 12 %, respectively. Maximum and minimum temperatures outside were 23 and 15 °C. In MASP, the light and scattered rain events were reported from 8 to 9 May and more intense after 12 May (weather station of University of São Paulo, IAG/USP). Inside (JQ) the tunnel, the registered conditions for temperature and relative humidity were 25.1 ± 0.6 °C and 38 ± 1 %, respectively, which shows more steady conditions. All the data regarding the meteorological variables are presented as supplementary material (Supplementary Tables S1 and S2). During the experiment inside (JQ) tunnel, the hourly average number of LDVs was 2247 vehicles h−1, from 7:00 am to 9:00 pm, being the maximum number observed in the evening at 7:00 pm, exceeding 3200 vehicles h−1 (Fig. 1). It is interesting to notice the high number of motorcycles from 7:00 am to 9:00 pm, reaching 550 units h−1 at 07:00 pm. The HDVs, urban cargo vehicles up to 6.3 m long, presented the lowest
ISORROPIA—thermodynamic aerosol model The ISORROPIA thermodynamic aerosol model resolves chemical equilibrium calculations among inorganic species with a high efficiency and rigorous modules for using in regional or global models. The ISORROPIA has been employed in several studies to compare/evaluate in situ datasets (Bian et al. 2014; Li et al. 2013; Livingston et al. 2009; Tao et al. 2014). The model approaches the activity coefficient calculation, algorithms, and solved equations, while further details were discussed previously (Fountoukis and Nenes 2007; Nenes et al. 1998). In this study, the version 2.1, often referred as ISORROPIA II, was applied to simulate the urban fine aerosol composition. ISORROPIA II can solve two types of problems: (i) forward and (ii) reverse. The forward problem is applied to solve a situation in which the aerosol precursors (gases) are known, whereas the reverse problem utilizes the aerosol phase concentration of NH3 (as ammonium salts or ammonium ion), H2SO4, Na+, Ca2+, K+, Mg2+, HCl, and
Fig. 1 Average hourly variation of vehicle fleet inside (JQ) urban tunnel in MASP, 4–14 May 2011. Vehicle types: light-duty vehicles (LDVs), heavy-duty vehicles (HDVs), Motorcycles, and CNG. Dashed lines represent the confidence interval (0.95)
Environ Sci Pollut Res
number of vehicles per hour (Fig. 1). Considering the vehicular fleet category profile running inside this tunnel and the fuel consumption in the same period (Brito et al. 2013), the gasoline and ethanol were the main fuels burned. Particulate matter mass concentration Figure 2 depicts the PM2.5 and PM2.5–10 mass concentration observed in both sampling sites. The relative maxima in PP for PM2.5 were observed at (N) period, which is reasonable due to intense traffic from 5 pm to 9 pm and because of the nocturnal boundary layer lowering. On the other hand, in the JQ site, the minimum mass concentrations were observed at night, during lighter traffic. The coarse fraction in PP showed a steep decrease in the mass concentrations after 7 May, on weekends and after light rain events. The average fine and coarse PM mass concentrations were 25.1 and 27.2 μg m−3 at (PP), respectively, whereas inside (JQ), they were 44.5 and 66.6 μg m−3. The median ratios between coarse and fine fractions on weekdays were 0.94 and 1.48 for PP and JQ, respectively. The particle fraction ratios inside the tunnel (JQ) highlight the coarse fraction dominance pattern over the fine (Fig. 2). The PM mass variability was dependent on the weather and the traffic outside (PP) and inside (JQ) tunnel, respectively. The direct PM emissions from civil engineering activities were some of the substantial air pollutant sources in addition to vehicle emissions in the PP site. Water-soluble species in particulate matter In order to evaluate the particulate matter ionic composition, the ionic balance (the sum of the cations in relation to the sum of the anions, in μeq L−1) was calculated for fine and coarse fractions separately. Regarding the fine fraction, PP and JQ showed good linear fit with slopes of 0.80 and 0.89, respectively (R2 > 0.89) and both presented anion deficit, which
Fig. 2 PM2.5 and PM2.5–10 mass concentration measured inside (JQ) and outside (PP) the urban vehicular tunnel, 4–14 May 2011. On the x-axis, the following ordination was used: M for 08:00 am–2:00 pm, A for 2:00
might be due to the lack of carboxylates and carbonate/ bicarbonate measurements. On the other hand, in PP and JQ, the coarse fractions presented similarly weak linear fit (R2 < 0.40), being the slope 0.39 at PP and 0.50 at JQ. These results suggest that some anions responsible for this significant bias were not considered in the analysis. For instance, CO32− and HCO3− which were not quantified, are considered key species for PM2.5–10 aqueous phase ionic balance (Karanasiou et al. 2011). Thus, considering calcium as a counter ion for CO32−, the improvement on linear fit was observed in ionic balance which, in this case, for JQ, the new values were slope = 0.84 and R2 > 0.78. It is important to highlight that calcium was the dominant ion in PM2.5–10 at both sites (Table 1). Table 1 summarizes the average and standard deviation of PM mass and ionic concentrations in both sampling sites. Ammonium exists almost solely in the fine fraction, showing an average percentage (m/m) in the PM2.5 ranging from 18.4 to 19.8 % for JQ and PP, respectively. On the contrary, in the PM2.5–10, the ratio between ammonium and total ion mass ranged from 0.7 to 5.8 %. Even though ammonium represents a significant quota in the PM2.5, the sulfate showed the highest concentration in both sampling sites. It is important to highlight that the nitrate showed similar concentrations in both PM fractions and sampling sites. Sulfate, nitrate, and ammonium ions aggregated accounted for 77 % of total water-soluble species in PM2.5 for JQ and 76 % for PP. In PM2.5–10, these ions contributed to 43 % at PP and 39 % at JQ. The ions sodium and calcium showed different concentration profiles, being the former similar in both PM fractions for each site, while the latter presented higher PM2.5– 10 then PM2.5 for both sites. The calcium sources can be distributed in abrasive (soil/pavement) and civil engineering process. The sodium presented higher concentration than the chloride in PM2.5, suggesting the Na+ direct emission contribution from fuel burning by vehicles (Vieira-Filho et al. 2013).
pm–08:00 pm, D for 08:00 am–08:00 pm, and N for 08:00 pm to 08:00 am
Environ Sci Pollut Res Table 1 The mass and water-soluble species average concentrations (±standard deviation) of PM2.5 and PM2.5–10, outside (PP) and inside (JQ) the urban vehicular tunnel in MASP, from 4 to 14 May 2011
Mass Na+ NH4+ K+ Mg2+ Ca2+ F− −
Cl NO3− SO42− Total
PP (outside tunnel)
JQ (inside tunnel)
PM2.5
PM2.5–10
PM2.5
PM2.5–10
μg m−3 25.1 ± 17.9 0.52 ± 0.20 1.14 ± 0.71 0.37 ± 0.20 0.05 ± 0.02 0.33 ± 0.12
27.2 ± 16.0 0.63 ± 0.27 0.26 ± 0.26 0.13 ± 0.08 0.17 ± 0.06 1.23 ± 0.48
44.5 ± 10.8 0.36 ± 0.09 1.27 ± 0.69 0.39 ± 0.13 0.10 ± 0.03 0.57 ± 0.20
66.6 ± 21.1 0.37 ± 0.17 0.03 ± 0.11 0.15 ± 0.06 0.22 ± 0.07 1.45 ± 0.54
0.03 ± 0.04 0.10 ± 0.08 0.80 ± 0.52 2.42 ± 1.45
0.05 ± 0.05 0.35 ± 0.35 0.91 ± 0.53 0.75 ± 0.32
0.06 ± 0.09 0.12 ± 0.06 0.76 ± 0.53 3.27 ± 1.76
0.07 ± 0.04 0.24 ± 0.15 0.70 ± 0.30 0.91 ± 0.33
5.76 ± 2.80 4.48 ± 1.71 6.91 ± 2.55 4.16 ± 1.20
Bold values represent the major contributors for each fraction of PM mass concentration
Nitrogen species and water content in PM2.5—ISORROPIA II Figure 3 depicts mass balance of PM2.5 total mass simulated by ISORROPIA II for a reverse thermodynamic stable condition. In this output, we considered the relative distribution of aqueous and solid species and the PM2.5 water content in samples from JQ and PP sites. The distribution among inorganic compounds (solid + aqueous species) for the total fine particulate matter was more accurate on thermodynamic models due to the following: (i) Several chemical reactions among the species were taken into consideration; (ii) the water content of particulate matter was calculated based on
Fig. 3 PM 2 .5 simulation from ISORROPIA II regarding the representatives of aerosol mass inside (JQ) and outside (PP) the urban vehicular tunnel, 4–14 May 2011. On the x-axis, the following ordination
meteorological variables; (iii) phases of particulate matter could interchange, gas into particulate phase and vice versa (Fountoukis and Nenes 2007). The total percentage of inorganic species and water at PP region was noteworthy (Fig. 3), reaching maximum values around 90 % for N periods. Overall, the representativeness percentage was always higher at N period than D period, following the relative humidity variability. From the total PM2.5 samples at PP, 55 % attributes around half of the total mass to inorganic species, including water, which was a major component in this estimate. Contrariwise, the inorganic species at JQ were limited to 30 % of total PM2.5; there was also a dominance of solid species during the whole simulation period (Fig. 3). For instance, in some periods (7 May, 8 May, and 12 May), the simulation indicated pure inorganic solid aerosols. Additionally, for the same experimental setup, the elemental carbon (17 %) and organic compound (42 %) contribution was estimated by the chemical mass balance for PM2.5 total mass of the JQ fewer sampling sets (Brito et al. 2013). In order to verify the partition of nitrogen compounds— NH 4 + (aq) , NH 4 HSO 4(s) , (NH 4 ) 2 SO 4(s) , NH 4 NO 3(s) , and NH4Cl(s)—at PM2.5 for both sites, a reverse problem for stable conditions was simulated by using ISORROPIA II. The simulation for PP indicated that ammonia interaction with the gases HNO3(g), HCl(g), and H2SO4(g) was not favorable given the predicted low concentration levels of ammonium bisulfate, ammonium sulfate, ammonium nitrate, and ammonium chloride (Fig. 4). For instance, ammonium ion was the dominant species in this simulation for the whole PP campaign with 1.19 ± 0.73 μg m−3 average concentration, which was substantially close to the measured value, 1.14 ± 0.71 μg m−3 (Table 1). The simulated nitrogen species concentration distribution suggested that an interaction with water molecules might have been the route for ammonia in PP, which could
was used: M for 08:00 am–2:00 pm, A for 2:00 pm–08:00 pm, D for 08:00 am–08:00 pm, and N for 08:00 pm–08:00 am
Environ Sci Pollut Res
Fig. 4 PM2.5 simulation from ISORROPIA II regarding the nitrogen partition inside (JQ) and outside (PP) the urban vehicular tunnel, 4–14 May 2011. On the x-axis, the following ordination was used: M for 08:00
am–2:00 pm, A for 2:00 pm–08:00 pm, D for 08:00 am–08:00 pm, and N for 08:00 pm–08:00 am
be simplified by the chemical transformation: NH3(g) + NH4+(aq) + OH−(aq). H2O(aq) On the other hand, ammonium ion predicted at JQ (0.22 ± 0.38 μg m−3) was almost 6-fold lower than the measured value (1.27 ± 0.69 μg m−3), showing that the solid species (ammonium salts) might have had greater influence in the nitrogen compound distribution inside the tunnel. This distinct pattern corroborated with the results presented in Fig. 3 showed that in the PM2.5, 30 % was solid inorganic species and the water content was insignificant. Solid species as ammonium sulfate and ammonium nitrate average concentrations in JQ were 1.05 ± 1.84 and 0.72 ± 0.66 μg m−3,
respectively. Overall, JQ presented a more diverse distribution of nitrogen species than PP. In summary, these results highlight the importance of water content for the ammonium phase distribution (solid/aqueous). In addition, all solid species presented higher concentrations in JQ.
Table 2 Backward stepwise regression for the PM2.5 water-soluble species dataset Model
Variables Dependent
Outside (PP) tunnel 1 [NH4+] 2 3 4 5 6 Inside (JQ) tunnel 1 [NH4+] 2 3 4 5 6
Stepwise regression analysis Table 2 shows results of the backward stepwise regression analysis for the PM2.5 water-soluble species at PP and JQ. For this analysis, the [NH4+] was considered a dependent variable, whereas all the other species concentrations, relative humidity, and temperature were considered explanatory variables. Both regions (PP and JQ) required six iterations to
Stepwise backward Independent
All ions [Cl−] [Na+] [Mg2+] [Tm] [RHm]
AIC(i)
−61.72 −63.88 −65.54 −67.46 −68.58 −69.5
ΔAIC
R2**
p value
7.78 5.62 3.96 2.04 0.92 0
0.93 0.94 0.94 0.95 0.95 0.95
10−6 10−7 10−8 10−8 10−9 10−10
Table 3 Regression coefficients (β0, βi, and R2) for backward stepwise regression (AIC algorithm) for both sampling sites, outside (PP) and inside (JQ) the tunnel Stepwise—backward—AIC method Model
−111.14 −113.14 −115.13 −117 −118.39 −119.16
8.02 6.02 4.03 2.16 0.77 0
0.96 0.97 0.97 0.97 0.97 0.97
[NH4+]
−13
10 10−14 10−15 10−16 10−16 10−16
Statistic indexes are given for outside (PP) and inside (JQ) tunnel sampling sites
Model parameters
Dependent Independent β0 JQ (inside)
All ions [Cl−] [Ca2+] [K+] [Tm] [RHm]
Variables
PP (outside) [NH4+]
[SO42−] [Na+] [NO3−] [Mg2+] [SO4 2−] [NO3−] [Ca2+] [K+]
p-value: *
