An analysis of diurnal cycles in the mass of ambient aerosols derived from biomass burning and agro-industry

Authors


Abstract

[1] Strong diurnal cycles in ambient aerosol mass were observed in a rural region of Southeast Brazil where the trace composition of the lower troposphere is governed mainly by emissions from agro-industry. An optical particle counter was used to record size-segregated aerosol number concentrations between 13 May 2010 and 15 March 2011. The data were collected every 10 min and used to calculate aerosol mass concentrations. Aerosol samples were also collected onto filters during daytime (10:00–16:00 local time) and nighttime (20:00–06:00) periods, for subsequent analysis of soluble ions and water-soluble organic carbon. Biomass burning aerosols predominated during the dry winter, while secondary aerosols were most important in the summer rainy season. In both seasons, diurnal cycles in calculated aerosol mass concentrations were due to the uptake of water by the aerosols and, to a lesser extent, to emissions and secondary aerosol formation. In neither season could the observed mass changes be explained by changes in the depth of the boundary layer. In the summer, nighttime increases in aerosol mass ranged from 2.7-fold to 81-fold, depending on particle size, while in the winter, the range was narrower, from 2.2-fold to 9.5-fold, supporting the possibility that the presence of particles derived from biomass burning reduced the overall ability of the aerosols to absorb water.

1 Introduction

[2] An understanding of the physical and chemical properties of tropospheric aerosols is important for many reasons. Airborne particulates have been linked to adverse health effects, such as asthma and hypertension [Gouveia and Fletcher, 2000; Gusjar et al., 2010; Martins et al., 2002; Ostro et al., 2007], and can act as vectors for the transport of chemical species, including nutrients, trace metals, and other contaminants. Aerosol dry deposition, the rate of which is governed by particle size and morphology, is an important mechanism for the transfer of chemical species to terrestrial and oceanic surfaces [Zufall and Davidson, 1998], while the size distribution of particles influences the rate at which they are removed from the atmosphere by precipitation scavenging [Jennings, 1998]. Atmospheric particles influence visibility [Malm et al., 1994] and radiative fluxes [Horvath, 1998], and their ability to act as cloud condensation nuclei (CCN) [Ramanathan et al., 2001; Tang and Munkelwitz, 1993, 1994] affects cloud albedo and lifetime, as well as precipitation processes [Albrecht, 1989; Freud et al., 2008; Pincus and Baker, 1994; Twomey, 1977].

[3] The influence on aerosol mass concentrations of factors including diurnal fluctuations in emission source strengths and meteorological parameters such as wind speed and boundary layer conditions has been reported for urban and rural areas [Baxla et al., 2009; Stanier et al., 2004]. There is a large body of literature concerning the hygroscopic properties of (mainly) submicron aerosols [Seinfeld and Pandis, 1998; Swietlicki et al., 2008]. However, information is scarce concerning the influence on aerosol mass of the substantial diurnal humidity fluctuations that are characteristic of many tropical and subtropical regions. The existence of humidity-related diurnal cycles in aerosol mass has important implications for both geochemical cycling (dry deposition of atmospheric aerosols to terrestrial and water surfaces) and human health (respiratory and other symptoms related to aerosol inhalation).

[4] In South America, there have been no previous reports on this issue, and aerosol research has mainly focused on two regions. In Amazonia, emissions from biomass burning and biota have been the focus of research programs including AMAZE, LBA-EUSTACH, LBA-CLAIRE, and LBA-SMOCC, among others [Andreae et al., 2004; Graham et al., 2003; Guyon et al., 2003; Mircea et al., 2005; Pöschl et al., 2010]. In the larger cities, especially São Paulo and Rio de Janeiro, problems of particulate air pollution are commonplace [Companhia de Tecnologia de Saneamento Ambiental (CETESB), 2011; Diretoria de Informação e Monitoramento Ambiental, 2009]. In contrast, much less work has been undertaken in the smaller cities and the interior of the subcontinent (except the Amazon).

[5] The present study concerns aerosols produced during the dry and wet seasons in rural Southeast Brazil. Here, agricultural biomass burning has (until recently) been widespread, and there are substantial emissions of resuspended soil dusts [Allen et al., 2010]. Field measurements were made in the central region of the State of São Paulo. The State covers an area of 248,209 km2 and has a population of around 40 million [Instituto Brasileiro de Geografia e Estatística (IBGE), 2011]. The economy of the central region is based on agro-industrial activities that, together with road transport, are the main sources of trace gases and aerosols. For many decades, the burning of sugar cane (necessary for manual harvesting) during the drier winter period has resulted in high emissions of aerosols and gases in the rural areas [Allen et al., 2004; Arbex et al., 2007; Da Rocha et al., 2005; Lara et al., 2005]. Around 4.59 × 106 ha (45,900 km2) of sugar cane were harvested in São Paulo State in the 2010/2011 season [IBGE, 2011], equivalent to 51.5% of Brazil's total area of the crop (8.91 × 106 ha). In 2010/2011, Brazilian production of raw cane was 635 × 106 t and that of São Paulo State was 323 × 106 t. At the time of the measurements described here, around 40% of the cane was being harvested manually, which requires prior burning of the crop before workers enter the plantations [CETESB, 2012]. However, legislation (State Law no. 11.241/02) requires the eventual cessation of sugar cane burning by 2031. Meanwhile, an agreement signed between the ethanol/sugar industry and the São Paulo State government envisages a much earlier elimination of burning, by 2017.

[6] The climate of central São Paulo State is characterized by two distinct seasons: the humid summer, during which most of the annual precipitation occurs, and the drier winter, when biomass burning is widespread. The spring and autumn seasons are short and not well defined. The experimental approach adopted involved the continuous monitoring of size-distributed ambient aerosol mass concentrations during winter (dry, biomass burning) and summer (wet, nonburning) seasons. At the same time, particles were collected onto filters for subsequent analyses of chemical composition. The ability of the aerosols to absorb water was evaluated by comparisons of particle mass during periods of high (nighttime) and low (daytime) relative humidity.

[7] The statistical techniques of agglomerative hierarchical clustering (AHC) [Everitt et al., 2001] and principal components analysis (PCA) [Jolliffe, 2002] are widely used to interpret environmental data [Townend, 2009] and were adopted here to characterize the sources of aerosols during the two seasons.

2 Methodology

2.1 Field Measurements

[8] Two sets of field data were acquired. The first set was obtained between 13 May 2010 and 15 March 2011, using a TSI Model 8240 AeroTrak optical particle counter, together with the collection of aerosols onto filters for subsequent chemical characterization. The second set of data was collected between 09:00 14 June 2012 and 09:00 15 June 2012, using the AeroTrak instrument together with a TSI scanning mobility particle sizer (SMPS) in order to measure the number concentrations of particles as small as 14.3 nm (aerosols were not collected for chemical analysis during this period). All measurements were made on the campus of São Paulo State University (UNESP), in a rural location ~4 km southwest of the town of Araraquara (site coordinates: 21°48′ 50.26′′S, 48° 12′ 07.64′′W; 658 m above sea level) (Figure 1). Equipment was installed on the roof of a building, at ~5 m above ground level (agl).

Figure 1.

Map showing the location of the field measurement station in São Paulo State. The insert shows the view over sugarcane plantations to the northwest of the site, with instrument inlets and meteorological sensors in the foreground.

[9] The optical particle counter recorded aerosol number concentrations in six particle size bins: 0.3–0.5, 0.5–1.0, 1.0–3.0, 3.0–5.0, 5.0–10.0, and >10.0 µm. These measurements were made as an average over 2 min, acquired every 10 min, and the data were stored by the AeroTrak instrument, prior to offline calculation of mass concentrations. The sample flow rate was 28.3 L min−1.

[10] The SMPS system used to measure the number concentrations of particles smaller than 0.3 µm comprised a TSI Model 3080 classifier, a Model 3081 differential mobility analyzer (DMA), and a Model 3775 butanol-based condensation particle counter (CPC) operated in low flow mode (0.3 L min−1). The classifier inlet was fitted with an impactor with an orifice diameter of 0.0457 cm, and the DMA sheath airflow was 3.0 L min−1. The upper and lower detectable aerosol size limits of the instrument were 14.3 and 615.3 nm, respectively. Scans were performed continuously, every 10 min, using an up-scan time of 120 s and a retrace time of 15 s.

[11] Ambient air was transferred directly to the AeroTrak and SMPS instruments using 1.5 m lengths of 1.0 cm ID copper tubing, with connections made using nonconductive plastic tubing (supplied by TSI).

[12] It should be noted that only particles larger than 0.3 µm were counted by the AeroTrak instrument, while the SMPS could detect particles as small as 14.3 nm. The greatest contribution to particle mass was derived from particles larger than 0.3 µm (i.e., those in the size range measured by the AeroTrak instrument). It has been found that in rural areas, most of the mass of ambient aerosols resides in supermicron particles, with only a small fraction of the total mass in particles smaller than 0.1 µm [Jaenicke, 1998]. In the study region, the measurements made in June 2012 of aerosols in the size range from 14.3 nm to >10.0 µm (using the SMPS in combination with the AeroTrak instrument) showed that 1.6% of the total atmospheric aerosol mass was present in particles smaller than 0.3 µm in diameter (Table 1).

Table 1. Aerosol Mass Concentrations (µg m−3) Measured in Different Size Fractions During the Period 09:00 14 June 2012 to 09:00 15 June 2012a
 Size Fraction (Particle Diameter, µm)
 0.014–0.30.3–0.50.5–1.01.0–3.03.0–5.05.0–10.0>10.0
  1. a

    Particles <0.3 µm in diameter were measured using a TSI scanning mobility particle sizer (SMPS). Particles >0.3 µm in diameter were measured using a TSI AeroTrak optical particle counter. Aerosol mass was calculated from number concentrations, assuming the particles to be spheres with a density of 1.2 g cm−3.

Concentration (µg m−3)26.989.5126480254554105
Contribution to total measured mass (%)1.65.57.729.315.533.86.4

[13] The number concentrations of aerosols measured in each size bin were converted to mass concentrations assuming spherical particle morphology (volume = 4/3πr3) and a density of 1.2 g mL−1. This is a midrange value of the densities of ambient aerosols reported previously [Kannosto et al., 2008] but is a lower limit estimate of the densities of secondary aerosols and mineral dusts [Pitz et al., 2008]. It is the same as the experimental value obtained for sulfuric acid aerosols at relative humidity (RH) = 75% [Kelly and McMurry, 1992]. The densities of dilute solutions (<10% m/v) of the inorganic salts expected to be found in the aerosols of the study region (such as (NH4)2SO4, KNO3, NaCl, and KCl, among others) lie in the approximate range 1.00–1.07 g mL−1. In another recent study, measurements were made of the size distributions of secondary organic aerosols under simulated tropospheric conditions, using SMPS instrumentation [Müller et al., 2012]. The measured size distribution data were used to calculate particle mass concentrations, assuming a particle density of 1.25 g mL−1, as also reported earlier [Saathoff et al., 2009]. In the latter work, aerosol density was determined by comparing aerosol mass spectrometer data with measurements made using an SMPS system. In the present study, water-soluble organic carbon (WSOC) contributed, on average, 66.5% of the mass of the soluble chemical species (Cl, NO3, SO42−, Na+, NH4+, K+, Mg2+, Ca2+, and WSOC) (Table 2), which further supports the use of an estimated density value of 1.2 g mL−1 in the mass calculations.

Table 2. Descriptive Statistics for Concentrations (µg m−3) of Chemical Species
 ClNO3SO42−Na+NH4+K+Mg2+Ca2+WSOC
First Quartile0.0180.2210.6770.0410.6230.2000.0440.2595.329
Median0.0430.4440.9750.0871.2000.4030.1200.5389.155
Third Quartile0.1311.1361.9590.1402.1010.7300.2060.89016.817
Mean0.2090.9961.6350.1011.4270.7270.1620.69211.783
SD0.3981.2811.7830.0741.1141.0210.1690.5488.838

[14] Non-size-segregated aerosol samples were collected onto filters for chemical analysis, during daytime (10:00–16:00 local time) and nighttime (20:00–06:00) periods. Sampling was performed at a flow rate of 10 L min−1, using polytetrafluoroethylene (Whatman, 47 mm diameter, 1 µm porosity) and cellulose nitrate filters (Sartorius, 47 mm, 0.45 µm porosity) to collect aerosols for subsequent analyses of major ions and WSOC, respectively. The exposed filters were folded in half, placed into zip-seal plastic bags, and stored at −22°C until the analyses were performed.

[15] Ambient temperature and relative humidity were measured continuously using a combination probe connected to a data logger (Campbell Scientific models HMP50 and CR10, respectively), and the data were subsequently downloaded to a laptop computer using Campbell's PC200W (v.3.3) software.

2.2 Chemical Analyses

2.2.1 Soluble Ions (Cl, NO3, SO42−, Na+, NH4+, K+, Mg2+, Ca2+)

[16] Filters were extracted into deionized water, and the solutions were analyzed using a Dionex DX-120 ion chromatograph equipped with interchangeable CG12A/CS12A and AG12A/AS12A 4 mm column sets (for cation and anion analyses, respectively), a conductivity detector, and a self-regenerating suppressor. Dilute carbonate and sulfuric acid solutions were used as the eluents for anion and cation analyses, respectively. Details of the ion chromatography system have been published previously [Da Rocha et al., 2003].

2.2.2 Water-Soluble Organic Carbon (WSOC)

[17] High-temperature catalytic oxidation employing a Shimadzu TOC 5000A total organic carbon analyzer was used for the determination of WSOC, as described previously [Campos et al., 2007; Coelho et al., 2008]. The extracted sample was first centrifuged (at 3000 rpm), after which an aliquot was injected for measurement of total dissolved carbon. The same sample was acidified with 25% H3PO4 and purged with a CO2-free carrier gas to transfer the CO2 generated from the inorganic carbon species to the detector. The concentration of WSOC was calculated as the difference between the total carbon and inorganic carbon concentrations. Filter blanks were analyzed using the same procedure employed for the exposed filters.

2.3 Statistical Analysis

[18] Agglomerative hierarchical clustering (AHC) and principal component analysis (PCA) were implemented using Addinsoft XLStat v. 2012 software (EasyStat Soluções Estatísticas, Campinas, Brazil). The PCA procedure [Jolliffe, 2002] employed the Pearson correlation coefficient as index of similarity, the orthogonal (varimax) rotation transformation, and the Kaiser normalization. PCA was performed using the chemical data alone. AHC [Everitt et al., 2001] used the Pearson correlation coefficient to evaluate similarity, and agglomeration was by unweighted pair-group averaging. The data set used for the AHC procedure included both the chemical data and the mass concentrations. These analyses employed a total of 104 samples (n = 104). Periods when sampling or analytical failures resulted in substantial loss of either aerosol number concentration or compositional data were not considered. The data used are listed in Table  S1 in the supporting information.

3 Results and Discussion

3.1 Aerosol Chemical Component and Mass Concentrations

[19] A summary of the chemical data is provided in Table 2. Time series plots of potassium (which is commonly used as a biomass burning marker) and sulfate (used as an indicator of secondary inorganic aerosols) are illustrated in Figure 2. The influence of biomass burning can be seen, with potassium concentrations increasing in June and declining by October. The mean concentrations of K+ during the winter and summer periods were 1.25 and 0.23 µg m−3, respectively. There was also increased production of secondary aerosols during the biomass burning period, reflected in mean sulfate concentrations of 2.36 and 0.94 µg m−3 for winter and summer, respectively. In previous work, secondary material was found to be an important component of rainwater in the study region [Coelho et al., 2011], indicative of the cloud scavenging of these aerosols and in agreement with the present observations.

Figure 2.

Concentrations of sulfate and potassium in total suspended particulates (TSPs) collected in central São Paulo State between 13 May 2010 and 15 March 2011. No samples were collected during the periods 1 August 2010 to 14 September 2010 and 11 December 2010 to 12 January 2011.

[20] Figure 3 shows the aerosol mass concentrations calculated from the AeroTrak data for three size fractions (0.3–1.0, 1.0–5.0, and 5.0 to > 10.0 µm), averaged for the periods during which samples were collected for chemical analysis. In all size fractions, concentrations increased during the driest months (July–October) when biomass burning and soil dust resuspension were most intense, with the largest size fraction (5.0 to  > 10.0 µm) generally contributing most to the total mass. This reflects a substantial increase in emissions of soil dusts in the region due to the activity of a large fleet of agricultural machinery in plantations and on unsealed tracks.

Figure 3.

Calculated mass concentrations of aerosols in three size fractions (0.3–1.0, 1.0–5.0, and 5.0 to  > 10.0 µm), measured in central São Paulo State between 13 May 2010 and 15 March 2011.

3.2 Diurnal Cycles in Aerosol Mass Concentrations

[21] Diurnal cycles in the calculated mass concentrations of aerosols in all size fractions measured by the optical particle counter (0.3 to >10 µm) were observed on most days. This behavior, characterized by nighttime mass maxima and daytime minima, is illustrated for the PM10 size fraction in Figure 4a for January 2011 (summer rainy season) and in Figure 4b for July 2010 (winter dry season). The PM10 values were calculated from the aerosol mass concentrations obtained from the first five AeroTrak size bins (0.3–10.0 µm), adjusted by a factor of 1.016 to correct for the 1.6% of the total mass that was present in particles smaller than 0.3 µm (which were not detected by the AeroTrak instrument, as discussed in section 2.1). Although several very large spikes in PM10 concentrations can be observed for both seasons (Figure 4), the diurnal fluctuations in concentrations were dissimilar. In January, extremely high PM10 concentrations occurred at, or near, the midpoint of the nocturnal peak. In July, most of the spikes occurred in the early evening, coinciding with the time that burning operations were conducted in sugar cane plantations. In this case, the mass increases were probably related to the presence of biomass burning aerosols, while during the summer, the absence of nighttime biomass burning resulted in distinctly different aerosol chemistry. These seasonal differences are discussed in sections 3.6 and 3.7.

Figure 4.

Calculated aerosol mass concentrations (PM10 size fraction) during (a) January 2011 (summer season) and (b) July 2010 (winter season), showing nighttime peaks in concentrations. The tick marks along the x axis correspond to 00:00 (local time).

[22] The simultaneous deployment of the SMPS and AeroTrak instruments in June 2012 enabled observation of the aerosol mass increases in the size range from 14.3 nm to >10 µm. These measurements revealed that there were nighttime increases of the masses of aerosols throughout this size range (discussed in section 3.4).

[23] The possible influence on aerosol concentrations of diurnal changes in the planetary boundary layer (PBL) depth was considered using previously reported ground-based lidar (light detection and ranging) measurements in central São Paulo State. Although published measurements for the region are scarce, the available data indicate that in summer, the top of the PBL lies at approximately 1.5 km agl at night and at up to approximately 3 km agl during the daytime [Held et al., 2008]. In winter, only small diurnal fluctuations were observed in the PBL depth, which remained at approximately 2–3 km throughout the day [Urban et al., 2012]. There are, therefore, no substantial seasonal fluctuations in the PBL depth in the study region, although in summer (but not winter), the PBL depth at night can reduce to around half the daytime maximum. This was unable to explain the amplitude of the diurnal cycle in calculated aerosol mass concentrations (Figure 4).

[24] An important point to note is that although ambient aerosol mass concentration data are routinely collected by national environmental protection agencies, in their standard configurations, the instruments used are unable to detect changes in aerosol mass resulting from condensation of species initially present in the gas phase. Aerosol monitors such as the Rupprecht and Patashnick TEOM (tapered element oscillating microbalance) or beta-attenuation gauges (for example, Met One Instruments' BAM-1020 monitor or the Andersen beta gauge) [Gomišček et al., 2004] are designed to comply with specific guidelines, notably those given in the United States Environmental Protection Agency's Federal Equivalent Methods for the PM10 and PM2.5 size fractions (particles with diameters ≤10.0 µm and ≤2.5 µm, respectively). The devices are equipped with systems that either condition the sampled aerosol to a defined temperature and/or relative humidity or use a reference sample flow to compensate for changes in ambient temperature and humidity. In the Federal Reference Method (FRM), aerosol mass is determined gravimetrically, after conditioning of unexposed and exposed filters at a relative humidity of 35 ± 2% and a temperature of 22 ± 2°C [Rees et al., 2004].

[25] Given the above considerations, it is clear that diurnal fluctuations in aerosol mass concentrations resulting from the cyclical uptake and evaporation of volatile species, as described in the present work, could not be identified from existing national aerosol databases.

3.3 Influence of Chemical Composition on Aerosol Mass

[26] The relationships between the total calculated aerosol mass concentrations (AeroTrak data) and the concentrations of the different chemical species were investigated using linear regression analysis. This was applied to (a) nighttime periods when the peak value of total calculated aerosol mass was at least twice the preceding daytime minimum value (35 samples) and (b) the corresponding daytime periods when mass concentrations were low (34 samples). In this analysis, all data were normalized using the ratio XxXx, where X represents each individual value of species x, and ΣXx is the sum of the individual values of species x in the data set. Figure 5 shows an example of the regression analysis, applied to nighttime concentrations of SO42− and aerosol mass in the 1.0–3.0 µm size fraction (which showed the best correlation with sulfate).

Figure 5.

Linear regression obtained between normalized aerosol mass concentrations in the 1.0–3.0 µm size fraction and normalized SO42− concentrations for nighttime periods when a peak in aerosol number concentrations was observed.

[27] The Pearson correlation coefficients obtained for the relationships between the mass and chemical component concentrations during the nighttime period are provided in Table 3. The slopes of the regression lines (for cases where the correlation was significant at the 95% level) are also given in Table 3, where a high slope value indicates that the mass concentration (calculated from the AeroTrak measurements) increased rapidly as the concentration of the chemical component increased. The results show that increases in the mass of the 0.3–0.5 µm aerosol fraction were associated mainly with WSOC, while increases in the mass of the 0.5–1.0 µm fraction were associated with WSOC (which showed the highest slope value), NH4+, and SO42−. The association of WSOC with aerosol mass, especially in the 0.3–0.5 µm fraction, is indicative of the presence of primary particles from biomass burning since the highest concentrations of levoglucosan (a biomass burning marker compound) were also found in this size range [Urban et al., 2012].

Table 3. Values of the Pearson Correlation Coefficient (r) Obtained Between Calculated Aerosol Mass Concentrations in Six Size Bins and Chemical Species Concentrations for Nighttime Measurements During Periods When Aerosol Mass Concentration Peaks Were Observed
 Particle Size Fraction (µm)
 0.3–0.50.5–1.01.0–3.03.0–5.05.0–10.0>10.0
  1. a

    Values in bold type indicate significance at the 95% level (α = 0.05).

  2. b

    Values in italics are the slopes of the regression lines obtained between aerosol mass concentrations and chemical species concentrations (only values significant at the 95% level are shown).

Cl−0.1910.0840.2240.243−0.109−0.213
NO30.0540.2280.347a (0.389)b0.399 (0.237)−0.063−0.175
SO42−0.2970.401 (0.529)0.497 (0.610)0.464 (0.301)0.2230.089
Na+0.000−0.106−0.0660.2220.0950.044
NH4+0.2190.378 (0.532)0.2550.024−0.049−0.083
K+0.0200.2220.399 (0.302)0.484 (0.194)0.082−0.081
Mg2+0.0030.0670.2380.448 (0.216)0.127−0.029
Ca2+0.1790.2800.369 (0.447)0.471 (0.301)0.2800.140
WSOC0.397 (0.252)0.454 (0.643)0.441 (0.581)0.3260.1120.028

[28] Concentrations of the ions K+, Mg2+, Ca2+, and NO3 increased together with particle mass in the size range 1.0–5.0 µm, which reveals the presence of resuspended soil dusts and the scavenging of gaseous nitric acid by these particles. Only small nighttime increases were observed for the calculated mass concentrations of size fractions above 5 µm, which consequently did not show any correlation with the concentrations of chemical components.

3.4 Hygroscopic Particle Growth

[29] The mass concentration of aerosols in a given size range is influenced by the formation of particles in that size range, as well as by particle growth or loss of material to the vapor phase. The formation of secondary aerosol species and the uptake of water vapor by particles are both important processes that act to increase aerosol mass.

[30] The concentrations of most of the measured chemical species (except Na+ and Cl) were higher at night (Figure 6), with night/day concentration ratios varying between 1.17 (Ca2+) and 2.90 (K+). There were increases in the concentrations of both primary (K+, Mg2+, and Ca2+) and secondary (NO3, SO42−, and NH4+) species, as well as WSOC (which can have both primary and secondary sources).

Figure 6.

Nighttime/daytime ratios of chemical species concentrations.

[31] However, the magnitude of the nighttime increase in aerosol mass could not be explained by the increased mass of the chemical components. The maximum and minimum calculated aerosol mass concentrations for 16–17 January 2011 (Figure 4a), during the summer, are used here as an example. At the time of the nighttime peak, the mass increases were 15.2-fold (0.3–0.5 µm fraction), 81.4-fold (0.5–1.0 µm fraction), 63.1-fold (1.0–3.0 µm fraction), 5.5-fold (3.0–5.0 µm fraction), 5.7-fold (5.0–10.0 µm fraction), and 2.7-fold (>10.0 µm fraction), relative to the daytime mass minimum (Figure 7). Considering all size fractions, there was a 13.1-fold increase in overall aerosol mass.

Figure 7.

Nighttime aerosol mass increases (Minc) during summer (16–17 January 2011) and winter (4–5 July 2010), according to particle diameter (Dp).

[32] The amplitude of the nighttime increase in aerosol mass was smaller in winter. For example, during 4–5 July 2010 (Figure 4b), the nighttime mass increases were 6.9-fold (0.3–0.5 µm fraction), 9.5-fold (0.5–1.0 µm fraction), 2.3-fold (1.0–3.0 µm fraction), 2.5-fold (3.0–5.0 µm fraction), 2.4-fold (5.0–10.0 µm fraction), and 2.2-fold (>10.0 µm fraction). There was a 2.8-fold overall increase in aerosol mass (around 5 times lower than in the summer). In all size fractions, the mass increase was smaller than observed in summer.

[33] A possible explanation for the greater aerosol mass at night is the uptake of water by the particles under conditions of higher relative humidity. During the summer and winter periods, the ranges of measured nighttime maximum relative humidity were 87%–94% and 55%–95%, respectively; while the ranges of daytime minimum relative humidity were 46%–66% (summer) and 22%–63% (winter). The diurnal cycle in relative humidity was, therefore, favorable for hygroscopic aerosol growth at night during both seasons. However, the amplitude of the nighttime mass concentration peak in summer exceeded that in winter. Another point that merits attention is that the mass of particles in the 1.0–3.0 µm size fraction increased substantially, by around 60 times in the summer, but only by around 2.5 times in the winter. As can be clearly seen in Figure 7, there was not only much greater growth of particles in the <3.0 µm size range in summer but also a shift in the growth curve toward larger particles (within the <3.0 µm range). This could be explained by greater aerosol formation and/or hygroscopic growth in summer. In turn, increased hygroscopic growth could be the result of (a) higher humidity and/or (b) more hygroscopic aerosol due to different chemical composition. In summer, in addition to higher humidity, there was a greater relative contribution of secondary aerosols (discussed in section 3.7), which tend to be more hygroscopic than biomass burning aerosols (discussed in section 3.5).

[34] In June 2012, during the winter, the combined use of the SMPS and AeroTrak instruments enabled observation of the diurnal cycle in the masses of aerosols in the size range from 14.3 nm to >10 µm (Table S2 and Figure S1). Figure S1a shows the temporal behavior of the mass concentrations of aerosols in the size fractions 0.014–0.02, 0.02–0.05, 0.05–0.1, 0.1–0.2, and 0.2–0.3 µm, calculated from the SMPS number concentrations. Figure S1b shows the corresponding temporal behavior of the mass concentrations of the size fractions 0.3–0.5, 0.5–1.0, 1.0–3.0, 3.0–5.0, 5.0–10.0, and >10.0 µm, calculated from the AeroTrak number concentrations. The mass increases at the time of the nighttime peak, relative to the daytime minimum, for the different particle size ranges were 56.0-fold (0.014–0.02 µm fraction), 64.2-fold (0.02–0.05 µm fraction), 45.7-fold (0.05–0.1 µm fraction), 23.6-fold (0.1–0.2 µm fraction), 19.1-fold (0.2–0.3 µm fraction), 8.0-fold (0.3–0.5 µm fraction), 21.3-fold (0.5–1.0 µm fraction), 27.5-fold (1.0–3.0 µm fraction), 17.8-fold (3.0–5.0 µm fraction), 12.8-fold (5.0–10.0 µm fraction), and 12.0-fold (>10.0 µm fraction).

[35] The mass increases showed a bimodal distribution (Figure S2), with maxima at 0.02–0.05 and 1.0–3.0 µm. Greater nighttime mass increases occurred for the size fractions <0.1 and 0.5–5.0 µm, compared to the 0.1–0.5 and >5.0 µm size fractions. A possible explanation for this behavior could be differences in the chemical composition and hygroscopicity of aerosols in the various size fractions.

3.5 Influence of Aerosol Composition on Water Uptake Capacity

[36] The uptake of water by aerosols is determined not only by meteorological conditions (especially humidity) but also by the chemical composition of the particles. The inorganic compounds potassium chloride (KCl), potassium nitrate (KNO3), and potassium sulfate (K2SO4) are often found in biomass burning plumes [Posfai et al., 2003; Rissler et al., 2006]. As the aerosol ages, chloride is lost to the gas phase as HCl, and there is a shift toward potassium nitrate and potassium sulfate salts [Freney et al., 2009]. The individual compounds have high deliquescence relative humidity (DRH): 84.2% (KCl), 96% (K2SO4), and 93% (KNO3), at 298 K; and mixtures of potassium salts also deliquesce at relative humidity of 80%–90% [Freney et al., 2009]. In contrast, the DRH of the secondary species NH4NO3 and (NH4)2SO4 are considerably lower, at around 62% and 80.0%, respectively [Tang and Munkelwitz, 1993], and those of the partially neutralized ammonium salts NH4HSO4 and (NH4)3H(SO4)2 are in the regions of 40.0% and 69.0%, respectively [Seinfeld and Pandis, 1998]. This suggests that hygroscopic growth should be greater for aerosols rich in secondary ammonium compounds (rather than the biomass burning species mentioned above). In addition, lower temperature and higher humidity at night favors the particle phase in the thermodynamic equilibrium established between ammonium nitrate, ammonia, and nitric acid (equation (1)), resulting in the formation of ammonium nitrate aerosols that can grow rapidly at fairly low relative humidity [Ansari and Pandis, 1998, 2000].

display math(1)

[37] From the above discussion, it can be concluded that diurnal fluctuations in aerosol mass concentrations should be favored by the presence of secondary aerosols rather than biomass burning particles, as well as by a wider diurnal humidity range.

[38] Previous work in the study region has shown that although the concentrations of many chemical components of the aerosol increase substantially in the winter dry season, the size distributions of most species remain constant throughout the year [Da Rocha et al., 2005]. Interestingly, an exception was nitrate, for which a very large submicron mode was observed in the winter, but not in the summer. A coarse particle mode was observed in both seasons. It was demonstrated that while aerosols smaller than 1.8 µm were acidic, those >1.8 µm were basic, favoring the stability of carbonate ions. This means that in winter, excess nitric acid should be absorbed by basic particles, which are therefore transformed into nitrate salts. As a result, a compound such as K2CO3 (DRH = 43%) is converted to KNO3 (DRH = 93%), and the aerosol becomes less hygroscopic. Analogous processes have been reported previously for mineral dust aerosols [Sullivan et al., 2007].

[39] In the work of Da Rocha et al. [2005], the submicron nitrate mode coincided with that of potassium, confirming the presence of KNO3 derived from biomass burning during the winter, which would act together with lower humidity to diminish the potential for particle growth. On the other hand, in summer, the mineral aerosols retained more of their original hygroscopicity, since there was less HNO3 available for the conversion reaction. The mean concentration of HNO3 was 2 times higher in winter (0.59 ppb) than in summer (0.28 ppb).

[40] A final consideration is that the hygroscopic behavior of particles of mixed composition, containing inorganic species together with soluble organic compounds, can differ from that of single-component aerosols. There is no clear consensus in the literature, with both increases and decreases in DRH values having been reported for mixed aerosols, compared to those containing inorganic substances alone [Choi and Chan, 2002; Miñambres et al., 2010; Wu et al., 2011; Zhang et al., 2011]. The chemical measurements (Table 2) showed that WSOC contributed 66.5%, on average, to the total mass of the soluble aerosol species measured, so that the existence of internal mixtures of inorganic compounds and WSOC could potentially affect aerosol growth rates.

3.6 Origins of the Aerosols

[41] Relationships between the analyzed parameters were first investigated using AHC applied to the full data set (Figure 8a). This revealed similarity among the species WSOC, Ca2+, K+, and Mg2+ (group 1) and Cl, NO3, and SO42− (group 2), with Na+ and NH4+ showing weaker similarity with groups 1 and 2, respectively. These results are indicative of the existence of two main sources of aerosols, namely, primary emissions (group 1) and secondary aerosol formation (group 2).

Figure 8.

Dendrograms obtained using agglomerative hierarchical clustering (AHC) applied to aerosol mass concentrations (in six size bins) and chemical components for (a) all seasons (n = 104), (b) winter (n = 48), and (c) summer (n = 56). Similarity was evaluated using the Pearson correlation coefficient; agglomeration employed unweighted pair-group averaging. The chemical species corresponding to the different factors (F) extracted using principal component analysis (PCA) are indicated adjacent to the x axis.

[42] Two main clusters were also obtained for aerosol mass in the different size fractions, which separated aerosols in the finer fractions (0.3–3.0 µm) from the coarser aerosols (3.0 to  > 10.0 µm). This distinction can be explained by the presence of primary biomass burning aerosols and secondary particles in the fine size range and greater predominance of resuspended soil dusts in the coarse size range.

[43] Further insight into aerosol sources was obtained by the application of PCA to the chemical component data. An initial scree plot analysis revealed the existence of four main factors, which explained 84.6% of the variability. The PCA procedure was then implemented. The first factor extracted, F1 (Table 4a), with high loadings for K+, Mg2+, WSOC, and Ca2+, could be attributed to emissions from biomass burning activities, including combustion aerosols as well as soil dusts released during operations involving vehicles and machinery. The presence of WSOC in this group suggests that an important fraction of the soluble organic material was related to biomass burning emissions, as found previously [Decesari et al., 2006; Hays et al., 2002; Urban et al., 2012].

Table 4a. Principal Component Analysis of Aerosol Composition: Factor Loadings After Varimax Rotation for All Seasons (n = 104)a
 F1F2F3F4
 Percentage of Variability = 30.9Percentage of Variability = 21.6Percentage of Variability = 19.5Percentage of Variability = 12.6
  1. a

    F1: Biomass burning and resuspended dust; F2: Aged secondary aerosols; F3: Scavenging of acidic gases by mineral aerosols; F4: Marine aerosol. Values in bold type correspond to the species indicated in Figure 8a.

Cl0.1740.1550.8900.107
NO30.2800.6330.6580.079
SO42−0.2080.7590.4160.019
Na+0.1910.1770.0920.939
NH4+0.1600.8820.0110.211
K+0.8210.2290.341−0.020
Mg2+0.8280.1150.3890.143
Ca2+0.7290.0710.2310.397
WSOC0.8200.250−0.1580.107

[44] In the second factor (F2), high loadings for SO42−, NO3, and NH4+ revealed the presence of secondary aerosols produced by reactions involving gas phase precursors (including NOx, SO2, and NH3). In the third factor (F3), high loadings for Cl, NO3, and SO42−, together with moderate loadings for the mineral cations, could be explained by the scavenging of acidic gases (HCl, HNO3, and SO2) by resuspended soil aerosols. All of these species are emitted during sugar cane burning operations [Da Rocha et al., 2003]. The high PCA loading for Na+ in the fourth factor (F4) was probably due to the periodic transport into the study region of modified marine aerosols originating from coastal regions.

3.7 Seasonal Analysis of Aerosol Origins

[45] Seasonal differences in the sources of the aerosols were investigated by applying the AHC and PCA procedures to two subsets of the data. The winter (sugar cane harvest) season considered the period 13 May 2010 to 7 October 2010 (n = 48); samples allocated to the summer (nonharvest) season were collected between 15 October 2010 and 15 March 2011 (n = 56).

3.7.1 Dry Season (Winter)

[46] The AHC dendrogram obtained for the winter (Figure 8b) revealed the same two main clusters of chemical species obtained for the full data. These corresponded to primary emissions from biomass burning and soil dust resuspension (group 1: K+, Mg2+, Ca2+, and WSOC) and secondary aerosols derived from reactions involving SO2, NOx, HNO3, HCl, and NH3 (group 2: SO42−, NO3, NH4+, and Cl). Na+ showed no similarity with the other chemical species, which can be explained by the presence of stationary high-pressure systems over central Brazil during the winter months, with less frequent incursions of maritime air masses.

[47] For the winter period, there was evidence of modest similarity between group 1 and the group consisting of the aerosol mass concentrations, which could be explained by a greater influence of the source represented by group 1 (primary emissions from biomass burning and soil dust resuspension) during the dry season.

[48] The results of the application of PCA to the winter data were similar to those obtained for the full data, with extraction of four factors that explained 81.3% of the variability (Table 4b). The first factor (F1) explained 30.3% of the overall variability, confirming the important influence of the emissions from biomass burning and soil resuspension, in agreement with previous findings for the composition of rainwater in this region [Coelho et al., 2011]. Emissions from other sources (mainly road transport and industry) remain relatively constant throughout the year. During the winter, the factor associated with the scavenging of acidic gases by mineral aerosols (Table 4b, F2) explained a greater percentage of the total variability, compared to the full data (Table 4a, F3), indicative of a greater relative contribution from activities involving sugar cane burning and concomitant soil dust resuspension during the winter season.

Table 4b. Principal Component Analysis of Aerosol Composition: Factor Loadings After Varimax Rotation for Winter (n = 48)a
 F1F2F3F4
 Percentage of Variability = 30.3Percentage of Variability = 20.7Percentage of Variability = 18.9Percentage of Variability = 11.4
  1. a

    F1: Biomass burning and resuspended dust; F2: Scavenging of acidic gases by mineral aerosols; F3: Aged secondary aerosols; F4: Marine aerosol. Values in bold type correspond to the species indicated in Figure 8b.

Cl0.1040.881−0.021−0.019
NO30.1850.7350.5850.003
SO42−0.1860.5460.629−0.011
Na+0.086−0.0210.0240.991
NH4+−0.025−0.0080.9120.024
K+0.8580.2280.185−0.054
Mg2+0.8670.2840.0040.085
Ca2+0.7540.126−0.1730.167
WSOC0.765−0.3230.252−0.043

3.7.2 Wet Season (Summer)

[49] The AHC results for the summer period are shown in Figure 8c. The distribution of cluster groups was different to that obtained for the winter, with two main clusters separating the chemical components and the aerosol mass concentrations. The presence of all the chemical components in a single main cluster indicates the existence of sources that were spatially and temporally more uniformly distributed in summer, compared to the winter. In summer, biomass burning was mainly confined to smaller fires for weed control on waste ground, and soil dust resuspension was reduced due to less use of vehicles and machinery in plantations, together with more frequent rainfall.

[50] For the summer period, there was greater similarity between the 0.5–1.0 and 1.0–3.0 µm mass fractions and less similarity of these with the 0.3–0.5 µm mass fraction. This contrasted with the winter period, when there was greater similarity between the 0.3–0.5 and 0.5–1.0 µm fractions and less similarity of these with the 1.0–3.0 µm fraction. In the case of larger particles, during the summer, the mass of particles >10 µm showed less similarity with the 3.0–5.0 and 5.0–10.0 µm mass fractions, compared to the winter period. Possible explanations for these differences include the effect of a seasonal shift in the relative contributions of biomass burning and secondary aerosols, as well as the suppression of soil dust resuspension during the summer.

[51] The results of application of PCA to the summer chemical data are shown in Table 4c. Four factors explained 84.1% of the total variance, with the factor attributed to secondary aerosols (F1) contributing the largest fraction (32.6%) and the factors attributed to biomass burning (F2), marine aerosols (F3), and biogenic emissions (F4) contributing 21.3%, 17.2%, and 13.0% of the total variance, respectively. The evidence of the presence of marine aerosols (F3), together with a moderate loading for Na+ in the factor associated mainly with secondary aerosols (F1), suggests that in many cases, the air masses sampled had originated over the Atlantic Ocean, subsequently traversing urban regions, including the metropolitan region of São Paulo and its satellite cities. Here there is intense industrial activity (~2000 large industries) as well as a fleet of around 7 million road vehicles [CETESB, 2011]. In 2010, emissions of NOx and SO2 in the metropolitan region of São Paulo were estimated to be 8.4 × 104 and 9 × 103 t yr−1, respectively [CETESB, 2011], which is indicative of the potential for downwind formation of secondary aerosols.

Table 4c. Principal Component Analysis of Aerosol Composition: Factor Loadings After Varimax Rotation for Summer (n = 56)a
 F1F2F3F4
 Percentage of Variability = 32.6Percentage of Variability = 21.3Percentage of Variability = 17.2Percentage of Variability = 13.0
  1. a

    F1: Aged secondary aerosols; F2: Biomass burning; F3: Marine aerosol; F4: Secondary aerosols from biogenic emissions. Values in bold type correspond to the species indicated in Figure 8c.

Cl0.1080.5640.6470.199
NO30.6990.3730.364−0.014
SO42−0.9120.2120.1090.115
Na+0.435−0.0130.8370.115
NH4+0.9260.1780.1690.112
K+0.5550.7160.049−0.134
Mg2+0.2980.7650.1000.319
Ca2+0.3880.5100.4620.285
WSOC0.0780.1480.1600.944

[52] Although sugar cane burning is not practiced during the summer, deliberate fires are still used to control weed growth. There was, therefore, a continuing presence of biomass burning aerosols in the summer (factor F2), in agreement with the earlier report of the presence of biomass burning species in the region's rainwater during the summer [Coelho et al., 2011]. For the summer data, the moderate loadings obtained for Ca2+ in factors F1, F2, and F3 suggest that this cation was not exclusively associated with soil dust resuspension within the rural region. Additional sources of calcium include cement dusts emitted from urban construction-related activities, as well as the spreading of calcium carbonate onto fields in order to adjust the soil pH.

[53] Intense biological activity in summer, when the rate of plant growth is much higher than in winter, explains the high loading obtained for WSOC in F4. Particle formation following reactions involving biogenic volatile organic compounds is in agreement with the observation that biogenic emissions are a source of organic carbon in rainwater during the summer, but not the winter [Coelho et al., 2011].

4 Conclusions

[54] Evidence is presented for an association between diurnal cycles in the mass concentrations of ambient aerosols and the uptake of water by hygroscopic particles. Seasonal differences in the amplitude of the diurnal fluctuations could be explained by both differences in relative humidity and the lower hygroscopicity of biomass burning aerosols, compared to secondary aerosols. Statistical analyses using AHC and PCA showed that during the winter dry season, the most important sources of aerosols were biomass burning and soil dust resuspension, while secondary aerosols predominated in the summer.

[55] An important consideration is that a shift is occurring in the chemical composition of tropospheric aerosols in this region. This is because agricultural biomass burning is being phased out, while at the same time, an extensive fleet of harvesting machinery is being deployed in São Paulo State's sugar cane plantations. As a result, the relative contributions of resuspended soil dusts and secondary aerosols are expected to increase, while that of biomass burning emissions will decrease. The effects on regional climate, aerosol deposition, and potential implications for human health will require evaluation. The present work provides a baseline for future studies, the goal of which will be to suggest possible impacts of a major change in the chemical composition of the atmospheric aerosol in an economically important region of South America.

Acknowledgments

[56] The authors are grateful for the financial support provided by FAPESP (Fundação de Amparo à Pesquisa do Estado de São Paulo; process 2008/58073-5) and CNPq (Conselho Nacional de Desenvolvimento Científico e Tecnológico; process 311668/2011-9 and process 303263/2009-1). We also thank Dr. Gerhard Held (IPMet, UNESP, Brazil) for helpful discussions concerning the meteorology of the study region.

Ancillary