Inorganic Chemical Composition of Fine Particulates in Medium-Sized Urban Areas : A Case Study of Brazilian Cities

The aim of this study was to characterize the inorganic chemical composition of fine inhalable atmospheric particulate matter (PM2.5) in medium-sized cities in Brazil. These cities account for a significant proportion of the population and are growing at rates above the national average, thereby demonstrating the importance of carefully analysing the possible impact of such growth on air quality over the coming decades. In 2013 and 2014, this study collected PM2.5 samples from sites in the cities of Londrina and Maringa in two seasons: winter and summer. The mean concentration of PM2.5 ranged from 4.4 μg m and 3.7 μg m during the summer to 10.3 μg m and 8.0 μg m in winter in Londrina and Maringa, respectively. The analysis of the major water-soluble ions, nitrate, sulphate and chloride, showed that they accounted for between 16.5% and 35.1% from the mass of PM2.5, with sulphate providing the greatest contribution in all the campaigns. The nitrate/sulphate ratios ranged from 0.2 and 0.6, which are similar to the figures cited in the literature for other regions of the world. Although the PM2.5 concentrations are much lower than those observed in Asia and in Sao Paulo the participation of ions (%) is very close to that observed in Asian cities and significantly higher than the values recorded in other areas of Brazil, possibly as a result of the increased influence of burning of biomass and waste. The metals Zn, Pb, Cu and Mn found in the samples from all campaigns indicate that, in general, mobile sources are the main contributor to PM2.5. The winter campaigns showed the highest concentrations of black carbon equivalent (BCe). Absolute principal component analysis and enrichment factor analysis indicate the contribution of vehicular emission sources and biomass and waste burning to the inorganic chemical composition of PM2.5.


INTRODUCTION
Cities concentrate a large proportion of the world's population.In most Latin American countries over 85% of the population is already living in urban areas (UN, 2014).This accentuated urban growth has occurred without urban planning, which has led to poor air quality, constituting a public health problem.Some large cities such as Mexico City, Sao Paulo and Santiago have a long history of monitoring air quality, while others have little data (UNEP, 2010).There is a lack of observational data across South America, especially in medium-sized cities.These cities are involved in almost the same polluting activities as the megacities, so nothing justifies the lack of studies in these cities.The profiling of fine atmospheric particulate matter (PM 2.5 ), for example, constitutes extremely important information about what could very likely be causing serious health problems for populations living in these medium-sized cities.
The fine particles emitted directly or formed by chemical reactions and physical processes vary significantly in concentration and chemical composition over space and time.They are important because they act directly and indirectly on the planet's radiation balance and consequently on its climate (IPCC, 2013).Furthermore, they are harmful to health due to their size and possible chemical composition (Brown et al., 1950;Pope III et al., 1991;WHO, 2006;Arbex et al., 2012;Kim et al., 2015).Globally, the main sources of PM 2.5 are the burning of fuels such as coal, oil, gasoline, diesel and biomass in general and gas-to-particle conversion reactions (Seinfeld and Pandis, 2006;Thorpe and Harrison, 2008;Pant and Harrison, 2013;Varrica et al., 2013).
Black Carbon (BC) is one of the most significant components of fine atmospheric particulate matter in terms of health, especially in relation to cardiovascular risks (Janssen et al., 2012;Schwartz and Lewis, 2012).According to Gerlofs-Nijland et al. (2012), BC is not a directly toxic component, but functions as a universal carrier for a wide variety of compounds because, according to Janssen et al. (2012), the BC particles have a higher infiltration rate into the body.Another important feature of BC is its capacity to absorb light in the visible range, therefore reducing the albedo of clouds, snow and ice, causing atmospheric warming and surface cooling leading to effects on solar radiation and climate (Bond et al., 2013;Petzold et al., 2013).
Sulphate (SO 4 2-) and nitrate (NO 3 -) ions are among the principal components of particulate matter (Dawson et al., 2014).Most of the times, these ions are formed during gasparticle conversion reactions, particularly the oxidation of SO 2 and NO x (Lin, 2002;Seinfeld and Pandis, 2006;Leibensperger et al., 2012) and neutralization reactions of sulfuric acid (H 2 SO 4 ) and nitric acid (HNO 3 ) in the presence of ammonia (NH 3 ).The sulphate ions may also originate from primary marine aerosols and soil particles.The NO x oxidizes in the aqueous phase and the formed HNO 3 may dissociate in water droplets, forming NO 3 -in the particulate phase (Seinfeld and Pandis, 2006;Allen et al., 2015).The presence of chloride ions (Cl -) in the particulate matter may occur as a result of the burning of biomass and incineration processes as well due to the contribution of the long-range transport of marine aerosols (Pelicho et al., 2006;Mihajlidi-Zelic et al., 2015).
Samples of particulate matter, originated from various sources, can show over forty trace elements such as metals.According to Sanders et al. (2003) and Silva et al. (2010), Cu is used as a lubricant additive and is present in the brake pads of motor vehicles and Pio et al. (2013) stated that Cu is a good marker of road traffic emissions.According to Silva et al. (2010), Zn is generally used as an anti-wear and antioxidant additive in engine oil.In Sao Paulo, Andrade et al. (2012) determined that elements such as Cu, Zn, S, Pb and Br in PM 2.5 are associated with vehicle emissions.According to Duan and Tan (2013), vehicles can directly emit Zn, Cu and Pb, together with exhaust gases and/or due to braking.The burning of fuels with additives results in emissions of Pb and Mn.The world's major urban areas have been the focus of studies of air quality and its impacts on health for a number of years, since they contain a significant proportion of the global population and PM 2.5 emission sources, mainly from vehicles.However, in large and developing countries like Brazil, China and India urban growth has been highly accelerated, especially in medium -to large-sized cities, compared with the growth rates of cities in Europe and North America.In Brazil, for example, medium-sized cities have grown sharply over recent decades, based on dataset provided by Brazilian Institute of Geography and Statistics (IBGE, 2015).Cities with populations of between 70,000 and one million inhabitants represent about one third of the Brazilian population and grew at an average rate of 1.7% per year between 2000 and 2010, compared to 1.2% annual growth in cities with more than 2 million inhabitants, including Sao Paulo.In addition, profiling of fine particulate matter from studies in developing medium-sized urban areas is scarce in Brazil.
Considering this situation, the aim of this study was to profile particulate matter in two medium-sized cities in the South region of Brazil.Both cities have witnessed major urban and population growth in recent decades.To profile the fine particulate matter, mass concentrations of black carbon, nitrate, sulphate, chloride and metals were analysed.It is important to highlight the lack of measurements in Brazil and South America in medium-sized cities and the fact that these are fundamental for the evaluation of local, regional and global-scale air quality models for the region.This type of information can have a direct impact on reducing uncertainties in future projections of the impacts of human activities on climate and health.Furthermore, the results may also provide supporting information for verification of emissions inventories for South America, where there is a large gap in detailed knowledge of emissions, primarily of aerosols.

Sampling
The study took place at sites located in Londrina and Maringa, cities with populations of about 550,000 and 400,000 respectively.The main economic activities of these cities are agribusiness, commerce and services.The vehicle fleet in these cities is also significant, with about 350,000 and 300,000 vehicles in circulation in Londrina and Maringa, respectively (IBGE, 2015).
In Londrina, atmospheric particulate matter was collected at the campus of the Federal Technological University of Paraná (UTFPR), located at 23°18"26.49"S,51°06'53.48"W,and with elevation of 557 m.This is a residential and agricultural area with traffic of light and heavy vehicles.In Maringa, atmospheric particulate matter was collected at the campus of the State University of Maringa (UEM), located at 23°24"18.74"S,51°56'17.43"Wand with elevation of 527 m.This other region, consisting of residential and commercial areas and near to major thoroughfares with heavy traffic of light and heavy vehicles.This study used meteorological data from National Institute of Meteorology (INMET) and UTFPR weather stations (Fig. 1).
We collected PM 2.5 samples in the years 2013 and 2014, during the winter and summer campaigns.In Maringa, sampling was performed in 2014 from 02/01 to 03/04 (summer) and from 07/02 to 08/02 (winter).In Londrina, sampling occurred in 2013 from 05 to 09/22 and from 11/27 to 12/29 (summer), and in 2014 from 08/03 to 09/03 (winter).The samples were collected over 24-hour periods, as described in previous studies (Fang et al., 2003;Andrade et al., 2012;Martins et al., 2012;Varrica et al., 2013), during 30 consecutive days at each collection point for each campaign, except for the winter 2013 campaign in Londrina for which there were only 18 days of sampling.

Sampling Preparation
Filters with PM 2.5 were divided in the middle and each part used for metal and ions analysis.By using a scalpel, we divided all filters, including the blank filters and those used to evaluate recovery.
All solutions were prepared using high purity reagents and ultrapure water (Purelab Option-Q, Elga, England).For the analysis of ions, atmospheric particulate matter was extracted using 1.0 mL of deionized water with maximum resistivity of 0.5 μS (Deionizer USF ELGA, England) (de Marques Freitas and Solci, 2009;Huang et al., 2013).Extraction time for each sample was 50 minutes, using a sonicator (USC-1600, Unique, Brazil, Indaiatuba), and the extract was filtered with a PTFE membrane with 0.22 μm porosity (Millex, Germany, Darmstadt).
To extract the metals, this study adopted a filter lixiviation process, using 65% of double-distilled nitric acid (Fluka).We establish the purity level for the analysis of the trace elements according to Fang et al. (2003) and Huang et al. (2013), using the microwave assisted digester (Titan MPS, Perkin-Elmer, USA).We also followed the Reference Notebook of Microwave applications for glass and quartz, adapted to the following conditions: step 1, ramp to 190°C, 35 bar at 80% of power for 30 min, step 2, 50°C, 35 bar at 0% of power for 10 min.(https://www.perkinelmer.com/lab-solutions/resources/docs/APP_Titan-MPS-Appl ications-Notebook-Industrial_010773B_01.pdf).

Analysis of Chemical Composition
After the extraction for ion analysis, an ion chromatograph (S-1100, Sykam, Eresing, Germany), equipped with a conductivity detector and a suppression system was used for the analysis of ions.The columns used were Dionex IonPac AS11 (4 × 250 mm) for separation, IonPac AG11 (4 × 50 mm) pre-column and the micro-membrane suppressor Dionex AMMS-II.The eluent used was a solution of NaOH 14.4 mmol L -1 with a flow rate of 1.5 mL min -1 .We used a solution of H 2 SO 4 9.0 mmol L -1 with a flow rate of 2.0 mL min -1 to regenerate the suppressor system and injected a volume of 50.0 µL for standards and samples, keeping the columns and conductivity cell at 35.0°C.The Seven anion Standard II (100.0 mg L -1 ) (Dionex) was used for the identification and quantification using calibration curves (de Marques Freitas and Solci, 2009).
Metals were analysed with a mass spectrometer with inductively coupled plasma (ICP-MS, NexION, Perkin-Elmer, Shelton, CT, USA), equipped with a sampling system with a peristaltic pump, a nebulizer and a concentric camera.The argon gas is introduced by an equipment has a dynamic reaction/collision cell in order to avoid polyatomic interference.For all measurements using the ICP-MS, we used Argon with a purity of 99.99% (White Martins).Prior to the determination of metals, we carried out a semi quantitative analysis in order to verify the metals existing in the samples and found the following metals: Al, Cd, Cr, Fe, Pb, Sn, Mn, Cu and Zn.We used the Rh (10 µg L -1 ) internal standard to measure the ICP-MS.The blanks of filters indicated a high level of some metals (Al, Fe, Mn, Cr) in the filters, which resulted in high quantification limit of the method.These results suggest that the quartz filters are not adequate to analyse trace metals in low concentrations.Huang et al. (2013) also found the high level for Al, Ca, Mg, Na background and low levels of trace elements.
During the analysis of ions and metals, we conducted a recovery test with filters impregnated with a solution of a known concentration, placed in a desiccator for at least 24 hours to dry, and then subjected to the process of extraction/lixiviation and determination to calculate the recovery percentage.The recovery was in the range of 80 to 90% (Allen et al., 2001).
The concentration of Black Carbon was determined using the light reflectance method (Sánchez- Ccoyllo et al., 2008;Andrade et al., 2012;Lack et al., 2014;Hetem and de Fatima Andrade, 2016) using a reflectometer (Diffusion Systems Ltd.Model 43 (M43D), London, UK).The conversion of the reflected light is inversely proportional to the light absorbed, which is a function of the quantity of absorbent material present in the sample.As what is being measured may not be 100% Black Carbon, Petzold et al. (2013) recommended the use of the term Black Carbon equivalents (BCe).The calibration curve for converting the reflected light into the concentration of Black Carbon originated from a calibration curve obtained empirically by Hetem (2014).
This study used the Absolute Principal Component Analysis (APCA) and Enrichment Factor (EF) techniques to identify the sources of the emissions.We obtained the APCA of the matrix by the principal component analysis, where the eigenvectors are factors obtained from the symmetrical decomposition and positive definite correlation matrix and the eigenvalue is indicated by the amount of variance explained by the factor.The Varimax rotation of the APCA scores adopted preserved the independence of the factors that are a measure of the contribution of each source to each sample of PM 2.5 .This analysis permits the estimation of the contribution of each factor to explain the mass concentration of PM 2.5 (Kessler et al., 1992;Ccoyllo and Andrade, 2002).The minimum condition for the number of samples is N ≥ 30 + (n + 30)/2, where N is the total number of samples and n is the number of variables (Henry et al., 1984).The sources of the PM 2.5 elements, Black Carbon equivalents, chlorides, nitrates and sulphates were estimated using APCA and the components with weightings of equal to or greater than 0.7 were considered.
The EF technique provided an overall assessment of the crustal contribution to the elemental composition of PM 2.5 collected in the cities of Londrina and Maringa (Clements et al., 2014;Fomba et al., 2015;Lin et al., 2015;Zhu et al., 2015).The EF compares elemental concentrations found in particulate matter with concentrations of these elements found in the Earth's crust, calculated according to Eq. (1): where C is the concentration of the element in atmospheric particulate matter or in the crust, x is the element to be considered and Al is the reference element (Clements et al., 2014).Generally, Fe, Si and Al become reference elements, as they are common and abundant in the crust (Taylor, 1964;Molnar et al., 1993).This study used the concentrations of metals in the Earth's crust, obtained from Souza Junior (2009) and Figueiredo (2014).
When EF is greater than 10, it means that there is significant enrichment of the element in relation to a source from the crust; when below 0.7, it turns into a strong depletion in comparison to the composition of the crust.We considered the elements with EF between 0.7 and 10 similar to and within the error range of the reference source, which implies that these elements may have a similar source of origin (Fomba et al., 2015).
The NO 3 -/SO 4 2-ratio has also been used to indicate the contribution of mobile and stationary sources of sulphur and nitrogen in the atmosphere (Yao et al., 2002;Wang et al., 2006;Lai et al., 2007).However, we recommend caution, taking into account the main sources existing in the region.
A trajectory analysis verified the contribution of other regions to the fine atmospheric particulate matter sampled.This analysis involved the use of the HYSPLIT model (Hybrid Single-Particle Lagrangian Integrated Trajectory), a complete system for calculating trajectories of air masses (Draxler et al., 2013;Pandolfi et al., 2014;Ripoll et al., 2014;Takahama et al., 2014).The model runs interactively on the internet through the READY system available on the website http://ready.arl.noaa.gov/HYSPLIT.php(Draxler et al., 2013).
For the processing and analysis of data the simple arithmetic means, standard deviation and medians were calculated.Pearson's correlation linear coefficients verified whether there was a relationship between different pollutants and between meteorological factors and pollutants (Chaloulakou et al., 2003;Huang et al., 2013;Alghamdi et al., 2015).
During the analysis of the similarities between the means of PM 2.5 mass concentrations, the Shapiro-Wilk test (test W) was used to verify whether the data set displayed a normal distribution or not.For data that showed a nonnormal distribution, a nonparametric Mann-Whitney test was used (Huang et al., 2013).For data with a normal distribution the parametric, Student's t test was used.

RESULTS AND DISCUSSION
A comparison between the means of PM 2.5 mass (Tables 1  and 2) recorded during winter and summer in both cities shows that, in most cases, the means are similar.Therefore, there is no statistically significant difference between the PM 2.5 concentrations of the two urban areas but there are differences in the PM 2.5 concentrations between winter and Table 1.Minimum, maximum, median, mean and standard deviation of concentrations of PM 2.5 (µg m -3 ), BCe (µg m -3 ), ions (µg m -3 ) and metals (ng m -3 ) from Londrina.summer.It is typical for volumes of rainfall to be lower in this region during winter, according to Alvares et al. (2014) while summers are hot and rainy, resulting in differences in the removal of pollutants from season to season.In Londrina, except for the winter campaign in 2013 (rainfall of 60.2 mm, historical average for period of 58.6 mm), rainfall was below the historical average during the sampling period (summer of 118.0 mm, winter of 53.3 mm and the historical average of 228.7 mm and 61.5 mm, respectively).In Maringa, in both campaigns, rainfall was above the historical average for the sampling period (summer of 265.1 mm and winter of 120.1 mm, and historical average of 211.7 mm and 62.2 mm, respectively).

Londrina
In Londrina, the average relative humidity and temperature during the winter campaign in 2013 were 54.5% and 22.7°C (historical average of 67.2% and 20.6°C).During the summer campaign, also in 2013, they were 67.0% and 24.9°C (historical average of 72.7% and 23.8°C).In the 2014 winter campaign, they were of 59.9% and 20.2°C (historical average of 67.5% and 18.8°C).In Maringa, in the summer 2014, the relative humidity and temperature were of 64.8% and 26.1°C (historical average of 74.6% and 24.7°C) and in the winter campaign in 2014, they were 69.1% and 18.8°C (historical average of 66.3% and 18.3°C).In Londrina, the winter was dryer and warmer than in Maringa, considering the sampling periods.
The city of Londrina showed the highest PM 2.5 concentration mean values, with clear skies or with a few clouds during the campaigns.An important contribution to these mass values could be the presence of compounds such as sulphate and nitrate as, according to Dawson et al. (2007) and Isaksen et al. (2009), clearer days represent higher light intensity, favouring the process of oxidation and the formation of these species.Allied to this are the average temperature values, which in all of the campaigns in both cities, were above the average of the historical data series for the sampling period, also contributing to the formation of sulphate and nitrate.In the campaigns carried out during winter there was a strong positive correlation (r ≥ 0.6) between temperature and PM 2.5 .
Another factor that contributed to the higher mass concentrations in Londrina was the volume of rainfall during the three campaigns, which was lower than the historical average for the period.In Maringa, rainfall volumes were 25.2% above the historical average during the summer campaign and 92.2% above average in the winter campaign.Even so, as winter is a relatively dry season in the region, the volume of rainfall was about half of that observed during the summer campaign.The wet deposition of atmospheric particulate matter is an efficient process for the removal of particulates, since the vast majority of the species are incorporated into raindrops and removed from the atmosphere.Rainfall also moistens the soil, thereby avoiding the resuspension of the particulates.Furthermore, according to Miranda and Andrade (2002), in winter the particles have a greater residence time, allowing agglomerates to arise and thereby reducing the concentration of particle numbers.According to Dawson et al. (2007), the mixed layer tends to be lower in the winter than in summer, thereby increasing the concentration of airborne particulate matter, and changes in the precipitation rate affect the concentrations of PM 2.5 more strongly in summer than in winter.
In Londrina, the average wind speed was less than 4.0 m s -1 during the three campaigns.In Maringa the average speed was less than 2.2 m s -1 during the two campaigns, which is higher than the historical average for the period (1.3 m s -1 in summer and 1.4 m s -1 in winter).In the winter 2014 campaign in Londrina, the average mass concentration of PM 2.5 was the highest of the five campaigns (in both cities), with the lowest average wind speed of Londrina campaigns.According to Chaloulakou et al. (2003) strong winds allow pollutants to disperse and weak winds result in their accumulation.According to Dawson et al. (2007) changes in wind speed affect all species that make up PM 2.5 and the increase in the mass concentration of PM 2.5 occurs more strongly in populated and polluted areas.During all campaigns the correlation with wind speed was low (≤ 0.4) or non-existent, indicating the contribution of local emissions sources to the mass of airborne particulate matters (Chaloulakou et al., 2003;Huang et al., 2013).
The mass concentration of PM 2.5 may be even higher as you get closer to the center of Londrina as the INMET station recorded very low average wind speed (0.9 m s -1 ), which makes it difficult for pollutants to disperse.Lopes et al. (2004) collected PM 2.5 samples in the centre of Londrina and obtained greater mass values than those found in this study in both campaigns (winter 2002 and summer 2003).
The mean concentrations of PM 2.5 during the sampling campaigns were between 3.7 µg m -3 (summer in Maringa) and 10.3 µg m -3 (winter in Londrina).Daily mean values were lower than the maximum recommended by the World Health Organization (WHO) for 24 hours of PM 2.5 (25 µg m -3 ).The only exception occurred in the summer 2013 in Londrina when the value reached 26.8 µg m -3 (12-13/12/2013).In this case, it was the result of an atypical event associated with excavation and covering of the soil with crushed stone near the sampling site.It is important to note that in Brazil there is not yet any established standard for PM 2.5 .
The mean concentrations of BCe recorded in Londrina and Maringa were very similar, ranging from 1.6 µg m -3 to 1.9 µg m -3 , except for the summer 2013 campaign in Londrina (0.9 µg m -3 ) (Tables 1 and 2).The concentration of BCe depends on the intensity of the emissions sources, dispersion and long-range transport.Therefore, similar values may indicate similar sources and removal processes.However, in this work, it was not possible to separate the contribution of the vehicular emission from the biomass burning.Recently, results obtained for 2015 indicated a significant contribution of biomass burning to BC concentrations for the place (Alves et al., 2016).These average concentrations were higher in winter campaigns because the removal of BC from the atmosphere is controlled by seasonal variations and this removal occurs more slowly in winter, when the atmosphere is more stable (Kompalli et al., 2014;Gabbi et al., 2015).The mean values recorded in Londrina and Maringa are lower than those recorded in cities such as Rio de Janeiro (2.4 µg m -3 ) by Godoy et al. (2009) and Sao Paulo (3.26 µg m -3 ) by Almeida et al. (2014).
Sulphate showed the highest average concentrations of ions, followed by nitrate and chloride, with the highest concentrations observed during the winter campaigns (Table 1) when there were periods of clear days with low precipitation, which result in an increase in the concentration of sulphate and nitrate according to Tsai and Cheng (2004).The relative high values of chloride in this region, which is far from the ocean, indicated a contribution of biomass burning, mainly in Londrina and in the winter campaign 2013 (average of 0.14 µg m -3 ).In Maringa, in summer, the average concentration was 0.05 µg m -3 .Values for the average concentrations of ions recorded during the campaigns in Londrina and Maringa (combined sum of the three ions) were between 1.1 µg m -3 and 2.0 µg m -3 .They lower than those observed by Souza et al. (2014) in Sao Paulo and about one order of magnitude lower than those found in large urban conglomerates in Southern China, as measured by Lai et al. (2007).In terms of PM 2.5 mass% , sulphate, nitrate and chloride ions measured in Londrina and Maringa presented values between 16.5% and 35.1%, values close to the proportions found in China (around 30%) and well above the percentage of around 13% recorded by Souza et al. (2014) in Sao Paulo and Piracicaba, a midsized city located 140 km from Sao Paulo.
In Maringa, the NO 3 -/SO 4 2-ratios were 0.2 in summer and 0.5 in winter, indicating values lower than other studies in the literature (Table 3).But nitrate was quantified in few samples, and this ratio is not representative of the summer in Maringa.In Londrina, the ratios were different, with 0.6 in summer and 0.4 in winter, with higher mean values than in Maringa, but lower than those found in the large urban centres in Brazil.Except for the summer campaign in Maringa, all of the other campaigns included at least one day where the NO 3 -/SO 4 2-ratio was greater than 1.Table 3 shows the ratios NO 3 -/SO 4 2-obtained in other studies reported in the literature for other regions.
The findings for the NO 3 -/SO 4 2-ratios in Londrina and in Maringa are close to those findings in Shanghai and Beijing recorded by Yao et al. (2002) and Shanghai by Wang et al. (2006) and slightly lower than the findings in Sao Paulo in 2007 and 2008 (Table 3).The difference between medium-sized cities and Sao Paulo may be due to the participation of different emissions sources.Studies in Sao Paulo by Allen et al. (1995), Bourotte et al. (2007), Vasconcellos et al. (2010) and Souza et al. (2014) have shown a decrease in sulphate values and an increase in the NO 3 -/SO 4 2-ratio.They were attributed to a significant reduction in the quantity of sulfur in diesel in Brazil, falling from 13,000 ppm in the 1980's to 500 ppm in 2009, with S10 diesel (10 ppm) available from 2012 onwards for new heavy vehicles (ANP, 2015).In the USA, Dominici et al. (2015) also observed a reduction in the sulphate concentration and a benefit effect on health.Unlike in Brazil however, in the case of the United States the reduction in atmospheric sulfur is mainly due to the replacement of coal with natural gas as fuel for thermal power stations (de Gouw et al., 2014).
The NO 3 -/SO 4 2-ratio is an indicator of the relative contributions of mobile and stationary emissions sources to atmospheric sulfur and nitrogen, particularly in studies carried out in Asia (Arimoto et al., 1996;Yao et al., 2002;Xiao and Liu, 2004).The use of this indicator in the region seems reasonable considering that NO x emissions are much higher than SO x emissions from the combustion of gasoline and diesel fuel in vehicles.Whereas, for the burning of coal by stationary emissions sources, during which SO x emissions reach double those of NO x emissions, a reverse relationship takes place.However, we must consider that Asian countries rely a lot on coal-fired power stations compared to the insignificant dependence in the areas of this study and other previous studies in Brazil.As such, the sulphur identified in the fine particulate matter analysed in this study is mainly of vehicular origin.However the proportions (%) of NO 3 -and SO 4 2-identified in Londrina and Maringa are approximately double the values found in the central region of Sao Paulo.Therefore, possible suggestions for contributions from emission sources other than direct vehicular activity include the burning of waste, heavy fuel oil and tires that can occur on the outskirts of these cities, as witnessed in Londrina, beside industrial activities.
We determined concentrations of Al, Cu and Pb in every campaign and quantified elements such as Sn and Zn in some samples of the campaigns and identified the elements Fe, Mn and Cr elements in all campaigns, but their concentration values were below the QL.In Maringa, the summer campaign showed the highest average concentrations for most of the metals, while, in Londrina, the winter campaign showed highest concentrations, especially in the 2013 (Table 1).As mentioned before, the relative high level of some metals in the filters limited the quantification of metals.
On average, the concentrations of Pb were relatively high when compared with figures obtained from PM 2.5 samples collected from tunnels in Sao Paulo.The mean values (18.6 ng m -3 for the winter 2014 campaign in Londrina and 40.0 ng m -3 for the summer 2014 campaign in Maringa) are higher than those found in the tunnel in Sao Paulo (18.1 ng m -3 ) (Brito et al., 2013).Nevertheless, this study quantified Pb concentration only in a few samples.It is important to emphasize that, unlike the samples taken inside tunnels, atmosphere samples from the sites used in this study were subject to contributions from the various existing sources of metals in the study area (combustion of fuels such as heavy oil, various industrial processes, construction, burning of tires and solid urban waste, etc.).They may explain the differences in relation to other studies.In Higashi-Hiroshima in Japan, Sakata et al. (2014) found that 80% of the total Pb was concentrated in fine particles of atmospheric particulate matter and the main sources were the incineration of solid urban waste and burning of heavy fuel oil.In southern China, Zhu et al. (2015) found that Pb in PM 2.5 derive mainly from industrial emissions and traffic.
The concentrations and composition of PM 2.5 may also have included contributions from fires and vehicular and industrial emissions from other regions of the country.The analysis of the trajectories of air masses showed us that, during the sampling periods, air masses passed mainly through the South, Southeast and Mid-west regions (Figs.2(A), 2(C), 2(E), 2(G) and 2(I)), regions with large urban centers such as Sao Paulo and Rio de Janeiro.According to CPTEC/INPE (2014) the MODIS satellite recorded a large number of fires in regions through which the trajectories of air masses passed during the winter campaigns (Figs.2(B), 2(F) and 2(J)), which may have contributed to the PM 2.5 composition.However, in the analyses carried out it was not possible to identify a tracer for burning of biomass.
Fig. 3 shows the percentage contribution of BCe, chloride, nitrate, sulphate and metals in the mass concentration of PM 2.5 .
The concentrations of ions and metals contributed with more than 25% of the total mass concentration of PM 2.5 in all campaigns.The largest contribution of BCe to PM 2.5 occurred in the summer campaign in Maringa (43.2%) (Fig. 3), a figure well above the findings of Andrade et al. (2012) in Sao Paulo and Rio de Janeiro (30%).The daily values recorded in this campaign do not differ from the other campaigns but the PM 2.5 values were lower.During this campaign, it rained above the historical average for the period, with the rainfall evenly distributed over 14 days, which may have contributed to a lower PM 2.5 value that resulted in this higher percentage.Other important chemical species were not analysed (organic carbon, K + , NH 4 + ), which is a limitation of this work.From the APCA for Londrina and Maringa it was possible to identify two factors.For Londrina factor 1 explained 33.5% of mass, associated with vehicular emissions due to the presence of BCe and sulphate.Factor 2 explained 7.6% of mass, associated with the burning of biomass and waste due to the presence of nitrate and chloride ions.In Londrina, there were incidents of fires (mostly domestic) during all of the campaigns, especially the winter campaigns, contributing significantly to the composition of PM 2.5 on site.It is important to note that the Brazil forbids the burning of biomass and waste, but the practice is common in the outlying areas of cities (see photos of the location in the supplementary materials).
Unlike in Londrina, there were no fires observed near to the sampling point in Maringa in the daily field observations.Thus, factor 1 explained 46.2% of mass due to the presence of sulphate and BCe, associated with vehicular emissions, and factor 2 explained 8.6% of the mass due to the presence of nitrate and chloride, mostly associated with industrial activity and burning of waste.
This study used Al as a reference element to calculate   (2015) found values above 10 for Cu and Cr, indicating anthropogenic origin and that the Cu and Mn elements originated from vehicular wear by abrasion and Zn and Pb indicate fuel combustion origins.However, the significant differences between the locations make it difficult to assign the elements found in these cities to the same sources.
There was a strong correlation (r ≥ 0.7) between Al and Mn, Zn and Cu, Zn and Al, Cu and Cr and Mn and Cu in Londrina, and Pb and Cu in Maringa, which, according to Islam et al. (2015), indicates a common source.The results indicate the contribution of vehicular activity and other sources to the concentrations of metals found, for example, the burning of household and industrial waste, which can emit Cr and the resuspension of soil rich in Al.In Londrina, the relatively high Pb values, found in few samples, seemed to derive mainly from the burning of tires and waste heavy fuel oil, possibly occurring near to the collection site (see supplementary material).In Maringa these values were attributed to the burning of fuels because there were no recorded occurrences of fires near the collection site like in Londrina and because there was a strong correlation with Cu.
In Shanghai, Wang et al. (2000) verified the contribution of vehicle emissions to lead concentrations in atmospheric particulate matter, even 2 years after the prohibition of adding lead to gasoline.In Sao Paulo, Weiss et al. (2009) suggested that traffic emissions continue to be the main source of Pb due to the large vehicle fleet, even with low concentrations of this metal in gasoline and ethanol.

CONCLUSIONS
The PM 2.5 samples were collected in the cities of Londrina and Maringa in both summer and winter campaigns to develop the profile of the inorganic composition of fine particulate matters in medium-sized cities in southern Brazil, which, despite representing more than one third of the population, have no air quality monitoring, and measurements are scarce.
The winter campaigns showed the greatest concentrations of mass, ions and BCe, and the city of Londrina presented the highest mean values, mainly due to the meteorological conditions during the campaigns.Although the analysis of the trajectories of air masses indicated the transport of pollutants originating mainly from burning in the southeast of the country, no tracer concentration for this type of source was determined.
The analysis of APCA and EF indicates the prevalence of mobile emissions sources in the inorganic chemical composition of PM 2.5 .However, the low nitrate/sulphate ratios encountered and the relatively high Pb concentrations suggest a significant influence from the activities of burning waste heavy fuel oil and tires in the medium-sized cities studied.In addition, more measurements may be done to better address this issue.
Higher Education Personnel (CAPES) and Fundação Araucaria for the financial support.We would also like to thank INMET for the data provided and the State University of Maringa for providing the space to perform the sampling.The authors gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT.

Fig. 2 .
Fig. 2. Trajectory of air masses, where the figures A, C, E, G and I represent the trajectories of the air masses arriving in Londrina and Maringa, respectively, calculated by HYSPLIT and Figures B, D, F, H and J represent the fire outbreaks.

Fig. 3 .
Fig. 3. Percentage contribution of BCe, chloride, nitrate, sulphate and metals in the mass concentration of PM 2.5 .