Central Amazonia serves as an ideal location to study atmospheric particle formation, since it often represents nearly natural, pre-industrial conditions but can also experience periods of anthropogenic influence due to the presence of emissions from large metropolitan areas like Manaus, Brazil. Ultrafine (sub-100 nm diameter) particles are often observed in this region, although new particle formation events seldom occur near the ground despite being readily observed in other forested regions with similar emissions of volatile organic compounds (VOCs). This study focuses on identifying the chemical composition of ultrafine particles as a means of determining the chemical species and mechanisms that may be responsible for new particle formation and growth in the region. These measurements were performed during the wet season as part of the Observations and Modeling of the Green Ocean Amazon (GoAmazon2014/5) field campaign at a site located 70 km southwest of Manaus. A thermal desorption chemical ionization mass spectrometer (TDCIMS) characterized the most abundant compounds detected in ultrafine particles. Two time periods representing distinct influences on aerosol composition, which we label as “anthropogenic” and “background” periods, were studied as part of a larger 10 d period of analysis. Higher particle number concentrations were measured during the anthropogenic period, and modeled back-trajectory frequencies indicate transport of emissions from the Manaus metropolitan area. During the background period there were much lower number concentrations, and back-trajectory frequencies showed that air masses arrived at the site predominantly from the forested regions to the north and northeast. TDCIMS-measured constituents also show distinct differences between the two observational periods. Although bisulfate was detected in particles throughout the 10 d period, the anthropogenic period had higher levels of particulate bisulfate overall. Ammonium and trimethyl ammonium were positively correlated with bisulfate. The background period had distinct diurnal patterns of particulate cyanate and acetate, while oxalate remained relatively constant during the 10 d period. 3-Methylfuran, a thermal decomposition product of a particulate-phase isoprene epoxydiol (IEPOX), was the dominant species measured in the positive-ion mode. Principal component analysis (PCA) was performed on the TDCIMS-measured ion abundance and aerosol mass spectrometer (AMS) mass concentration data. Two different hierarchical clusters representing unique influences arise: one comprising ultrafine particulate acetate, hydrogen oxalate, cyanate, trimethyl ammonium and 3-methylfuran and another made up of ultrafine particulate bisulfate, chloride, ammonium and potassium. A third cluster separated AMS-measured species from the two TDCIMS-derived clusters, indicating different sources or processes in ultrafine aerosol particle formation compared to larger submicron-sized particles.
Atmospheric aerosols are ubiquitous in the troposphere, and organics
contribute a large fraction to their chemical composition (Jimenez
et al., 2009). Models continue to have difficulty estimating the organic
contribution to aerosols in regions with both biogenic and anthropogenic
influence (Shrivastava et
al., 2017). Anthropogenic emissions have increased with global population,
and the resulting influences of such emissions on secondary organic aerosol
(SOA) formation continue to be assessed (Hofmann, 2015). The
reactive chemistry of organics in the presence of different regulating
species from urban sources, like sulfur dioxide (
The Amazon Basin is an ideal location to study how biogenic emissions,
anthropogenic trace gases and oxidants, and biomass burning impact the
number and composition of atmospheric aerosol particles. The Amazon Basin is
one of the few remaining tropical regions on Earth in which near-natural
conditions, free of direct anthropogenic influence, can be found. It has
been referred to as the “Green Ocean”, since particle concentrations can
be as low as those seen over the ocean, and, like the marine atmosphere, small
changes in particle properties can have a major impact on clouds and climate
(Andreae et al., 2004). While
isoprene is the most abundantly emitted biogenic volatile organic compound
(BVOC), monoterpenes and sesquiterpenes are observed in amounts potentially
sufficient for influencing particle composition (Alves
et al., 2016; Jardine et al., 2011, 2015; Yáñez-Serrano et al.,
2015; Yee et al., 2018). While on an annual basis, aerosol particle sources
in the Amazon Basin are dominated by the oxidation of BVOCs by OH and
Aerosol properties in the Amazon Basin show a seasonal dependence,
reflecting seasonal variability in emissions and deposition. During the wet
season (December through March), the region is dominated by natural
emissions, as accumulation-mode (particle diameters between 0.1 and 2.5
Unlike in other forested regions, particles with a diameter smaller than 30 nm are rarely observed over the Amazon Basin, suggesting that new particle formation events seldom occur near the ground (Martin et al., 2010). In other regions, new particle formation has been seen to occur during the daytime under sunny conditions, suggesting that both boundary layer dynamics and photochemistry are important factors (Bzdek et al., 2011). Varanda Rizzo et al. (2018) recently analyzed 4 years of particle size distributions acquired at the TT34 tower site located 60 km northwest of Manaus. Regional new particle formation and growth events were detected in only 3 % of total days observed, whereas bursts of ultrafine particles that lasted as least an hour occurred on 28 % of the days. Those “burst events” were equally likely to occur during the daytime as during the night, and the authors hypothesized that daytime events were caused by interrupted photochemical new particle formation, whereas nocturnal events might be due to emissions and/or fragmentation of primary biological particles. Recent airborne observations in the Amazon suggest that particle nucleation and growth can be initiated in the upper troposphere, with upwelling air masses transporting reactants into the free troposphere and downwelling air masses transporting aerosol particles and condensable compounds back into the boundary layer, where particles can continue to grow via condensation and coagulation (Andreae et al., 2018; Fan et al., 2018; Wang et al., 2016). Once formed, ultrafine particles can be key participants in a variety of atmospheric processes. One example of this is the subject of a recent study by Fan et al. (2018), who have suggested that ultrafine particles can increase the convective intensity of deep convective clouds. High concentrations of ultrafine particles, when present with high water vapor concentrations that are typical in the Amazon atmosphere, can form high concentrations of small cloud droplets that release latent heat and thereby result in more powerful updraft velocities.
While recent research provides some clarity on the origin, transport and climate impacts of ultrafine particles in the Amazon, very little is known about the chemical composition of these particles. Globally, measurements show that a major component of atmospheric ultrafine aerosol is organic compounds produced from BVOC oxidation (Bzdek et al., 2011; Riipinen et al., 2012; Smith et al., 2008). Many of these direct measurements of the composition of atmospheric ultrafine particles have been performed using the thermal desorption chemical ionization mass spectrometer (TDCIMS; Voisin et al., 2003). For example, TDCIMS measurements performed outside of Mexico City attribute about 90 % of the growth of freshly nucleated particles to oxidized organics (Smith et al., 2008). In the boreal forest of Finland, the contribution of oxidized organics is close to 100 %, and an analysis of composition suggests that marine emissions can play an important role in that process (Lawler et al., 2018). For the smallest particles measurable by TDCIMS, with diameters from 8 to 10 nm, between 23 % and 47 % of the constituents may be derived from organic salt formation, a reactive uptake mechanism that requires the presence of strong bases such as gas-phase amines (Smith et al., 2010).
Similar to other parts of the world, particles in the Amazon Basin are typically composed of 70 %–80 % organics by mass in both the fine and coarse size ranges (Graham et al., 2003). The composition of ultrafine particles has not been directly measured, although one study has proposed that the major component could be oxidized organics that have condensed onto potassium salt-rich primary particles emitted from active biota (Pöhlker et al., 2012). An understanding of the origin and chemical composition of ultrafine particles in the Amazon gives insight into their formation and growth processes. To improve on modeling the coupling of chemistry and climate in this sensitive region, incorporating accurate representations of particle formation and growth processes is required.
The most recent, and currently the largest, field campaign for studying the Amazon atmospheric chemistry and cloud processes was the Observations and Modeling of the Green Ocean Amazon (GoAmazon2014/5) experiment, which took place outside of Manaus from 1 January 2014 to 31 December 2015 (Martin et al., 2016). Two intensive observational periods (IOPs) were carried out during GoAmazon2014/5, corresponding to wet and dry seasons in 2014. This paper explores the chemical composition of ultrafine particles observed by TDCIMS during IOP1, which took place from 1 February to 31 March 2014. Specifically, we focus on 10 consecutive days that experienced air masses from both remote, primarily forested regions and the large metropolitan region of Manaus. This study investigates the influence of anthropogenic and biogenic emissions on the chemical composition of ultrafine particles in this region, from which one can infer the chemical processes that led to the formation and growth of ambient ultrafine particles in this region. The time evolution of select compounds in ambient ultrafine particles is analyzed, and compared to AMS measurements, using principal component analysis (PCA) in order to gain additional insights into the contribution of various emission sources to ultrafine-particle composition.
All data presented were collected at the T3 site (3.2133
Ambient ultrafine-particle composition was characterized using TDCIMS.
TDCIMS is an instrument designed specifically for the measurement of the
molecular composition of size-resolved ultrafine aerosol particles (Smith
et al., 2004; Voisin et al., 2003). In brief, sampled atmospheric particles
are charged by a unipolar charger and are collected via electrostatic
deposition on a platinum (Pt) filament over varying collection times. During
this campaign, collection times were either for 1 h or 30 min,
depending on the anticipated sample mass. A typical sample mass collected on
the filament ranged from 10 to 100 ng. After collection, the filament was
moved into an atmospheric pressure chemical ionization source region and
resistively heated to desorb the particulate-phase components. These
desorbed components were chemically ionized and detected using a quadrupole
mass spectrometer (Extrel Corp.). A zero-air generator (Parker Hannifin,
model HPZA-3500) provided the source of reagent ions
Both positive- and negative-ion-mode chemical analyses were performed during the two IOPs and are publicly available on the campaign data archive (Smith, 2016). During IOP1, several days of measurements were impacted by intermittent power outages and brownouts. IOP2 was characterized by comparatively lower concentrations of ultrafine particles, which is consistent with prior observations (Martin et al., 2010; Varanda Rizzo et al., 2018). Because of this, we focus our analysis on 10 consecutive days during IOP1 when instruments were operating consistently. This period also happened to coincide with the arrival of two distinct and consecutive air masses, which allows for more accurate side-by-side comparison of aerosol properties during these periods.
Ambient particles were sampled through 3 m of Cu tubing with a 0.63 cm inside diameter. The inlet extended 0.5 m above the roof of the laboratory and was curved downward and covered with a screen to prevent rain and insects from entering. Ambient particles during GoAmazon2014/5 were not size-selected prior to collection on the filament because of low ambient concentrations. The collection process, however, is inherently dependent on particle mobility (McMurry et al., 2009). In order to determine the size-dependent collection efficiency, tests were run at the start of the campaign by generating and collecting ammonium sulfate particles in the diameter range of 8–90 nm. The size-dependent TDCIMS sampling collection efficiencies were used to determine the volume mean diameter and estimated mass of each sample, as described in Smith et al. (2004).
To complement the TDCIMS dataset, high-resolution time-of-flight aerosol
mass spectrometry (AMS; Aerodyne, Inc.) was used to characterize
non-refractory compounds in PM
PCA was performed using the “princomp”
function of the R statistical software package
(R, 2011). A hierarchical cluster analysis was
performed using Ward's averaging method in the “hclust” function in R.
Ward's minimum variance method of hierarchical clustering was used, which
groups species within the same cluster to minimize the total variance
(Wilks, 2011). The purpose of this
analysis is to identify species or groups of species that may have unique
sources, trajectories or other physicochemical characteristics. Cluster
analysis was done for the following TDCIMS negative- and positive-ion-mode
species:
Meteorological data from the T3 site, showing planetary boundary layer height (green), rainfall
(light blue), relative humidity (dark blue), temperature (red), wind direction (purple), wind speed (black)
and total number concentration of sub-100 nm particles (
The 10 consecutive days that are the focus of this study can be
characterized by two distinct air mass types, as determined from
meteorological data and AMS-derived positive matrix factorization (PMF) factors (de Sá et al.,
2018). The first period, referred to as the anthropogenic period, was
from 14 March to mid-morning on 19 March, and the second period, the
background period, was from mid-morning on 19 March to 24 March. The
AMS-derived biomass-burning factor (BBOA), associated with levoglucosan, and
the anthropogenically dominated factor (ADOA), associated with mass fragment 91 or
“91fac” (
Wind direction data shown in Fig. 1, as well as NOAA HYSPLIT data shown in Fig. 2, suggest a reason for the two distinct periods. Back trajectories show that air masses during the anthropogenic period either pass through Manaus or south of Manaus prior to arrival at the T3 site. During this period, air masses most frequently passed over the main roadway that connects Manaus with Manacapuru, a neighboring city with a population of 93 000. Along this roadside are homes, agriculture and brick kilns, all of which contribute to local gas and particle emissions. In contrast, during the background period, air masses arrived at the T3 site most frequently from the northeast and west. Air masses that were measured at the site typically originated from densely forested regions northeast to west of Manaus. Less frequent were periods where air masses reaching the site originated from the east and were influenced by the Manaus metropolitan area. For example, during the evening of 21 March there was a period of increased number concentration, and as winds were quite stagnant at night, it is possible that a local emission source could have impacted the site during that period. Wind direction on this day corresponded with air masses arriving to the T3 site from the Manaus area.
Estimated masses of ultrafine particles sampled by TDCIMS were
determined and compared for the two periods (Fig. S3). During the
anthropogenic period there was no distinct diurnal pattern observed, with an
average of
Back-trajectory frequencies performed using HYSPLIT, showing the different air masses that travel to the T3 site during the anthropogenic period and background period. For each period, 20 trajectories were used to determine integrated frequencies spanning the 5 d of each period (14–19 March for the anthropogenic and 20–25 March for the background period). Each trajectory duration was 72 h. The color scale indicates the frequency of which air masses pass over that area, with the warmer color indicating that the area is more frequently passed over.
The five most abundant negative ions, as observed in full mass spectra (Fig. S1) taken at the start of the wet season campaign, are attributed to
Figure 3a shows the trend in ion fraction for the five most abundant negative
ions and four most abundant positive ions during the 10 d period of
analysis. During the anthropogenic period, the observed bisulfate ion (
Interestingly,
Of the measured positive-ion species,
PCA was performed on TDCIMS and AMS measurements to provide insights into the possible drivers for ultrafine-particle formation. Figure 4 shows the results of this analysis. In these plots, positive correlations are shown in blue, while negative correlations are shown in red. The intensity of the color and eccentricity of the ellipse are an indication of the degree of correlation. Pale-colored circles (eccentricity approximately zero) show little to no correlation, narrow ellipses with a positive slope and darker blue color illustrate strong positive correlation, and narrow ellipses with a negative slope and darker red color show strong negative correlation.
Principal component analysis (PCA) of TDCIMS and AMS data. Refer to text for details on the
interpretation of these plots. Shown are PCA results in which species are grouped into hierarchical clusters, with clusters
outlined by weighted black lines. Species are ordered by decreasing correlation to the first principal component,
from the top to bottom. TDCIMS chemical assignments for fragments are
Hierarchical clustering of these measurements results in three main clusters
of related particle constituents. This represents a series of clusters where
the species within each cluster covary, therefore being indicative, in this work,
of similar particle characteristics, processes or sources. The first cluster,
labeled “Cluster 1” in Fig. 4, grouped TDCIMS-derived cyanate (
With respect to PCA performed on the two datasets, Cluster 1, which includes
TDCIMS fragments typically linked to organic species (
Hierarchical clustering separates TDCIMS-measured ions into two clusters,
with Cluster 3 including TDCIMS-derived bisulfate (
The chemical composition of ultrafine particles in the Amazon Basin, as
measured during the GoAmazon2014/5, has two distinct influences: sources and
processes linked to anthropogenic origin and those related to more natural
sources and processes. During periods of heavier anthropogenic influence,
higher number concentrations of sub-100 nm particles were observed (Fig. 1). HYSPLIT back trajectories during the anthropogenic period (Fig. 2) not
only intersect with the Manaus metropolitan area but also with the main roadway
that connects Manaus with the city of Manacapuru. Influence from
anthropogenic sources, which during the study period is primarily linked to
Manaus metropolitan area emissions, may continuously affect the composition
of ultrafine particles observed at the T3 measurement site. Particulate
sulfate, measured as the bisulfate ion, was an important and dominant
contributor to the TDCIMS ion fraction during the anthropogenic period (Fig. 3) but was still measured, to a lesser extent, in the background period,
suggesting an omnipresent influence. The most abundant negative-ion species
measured during this campaign, likely related to organic nitrogen species at
Campaign datasets for Observations and Modeling of the Green Ocean Amazon (GoAmazon) are available at
The supplement related to this article is available online at:
JNS, PA, STM, OVB, RdS and JT designed the measurement campaign, and JNS, MJL, JO and SSdS carried out measurements. HSG performed data analysis, assisted by JNS and AC. HSG prepared the paper, with contributions from all co-authors.
The authors declare that they have no conflict of interest.
This article is part of the special issue “Observations and Modeling of the Green Ocean Amazon (GoAmazon2014/5) (ACP/AMT/GI/GMD inter-journal SI)”. It is not associated with a conference.
Institutional support was provided by the Central Office of the Large Scale Biosphere-Atmosphere Experiment in Amazonia (LBA), the National Institute of Amazonian Research (INPA), and Amazonas State University (UEA) and the Fundação de Amparo à Pesquisa do Estado do Amazonas (FAPEAM). We acknowledge support from the Atmospheric Radiation Measurement (ARM) Climate Research Facility, a user facility of the United States Department of Energy, Office of Science, sponsored by the Office of Biological and Environmental Research, and support from the Atmospheric System Research (ASR; DE-SC0011122 and DE-SC0011115) program of that office. James N. Smith acknowledges support from a Brazilian Science Mobility Program (Programa Ciência sem Fronteiras) Special Visiting Researcher scholarship. Paulo Artaxo acknowledges funding from FAPESP – Fundação de Apoio à Pesquisa do Estado de São Paulo, grant numbers 2017/17047-0, 2013/05014-0 and 2014/50848-9.
This research has been supported by the DOE Atmospheric System Research (grant nos. DE-SC0011122 and DE-SC0011115).
This paper was edited by Lynn M. Russell and reviewed by three anonymous referees.