The occurrence of nonliquid and liquid physical states of submicron atmospheric particulate matter (PM) downwind of an urban region in central Amazonia was investigated. Measurements were conducted during two intensive operating periods (IOP1 and IOP2) that took place during the wet and dry seasons of the GoAmazon2014/5 campaign. Air masses representing variable influences of background conditions, urban pollution, and regional- and continental-scale biomass burning passed over the research site. As the air masses varied, particle rebound fraction, an indicator of physical state, was measured in real time at ground level using an impactor apparatus. Micrographs collected by transmission electron microscopy confirmed that liquid particles adhered, while nonliquid particles rebounded. Relative humidity (RH) was scanned to collect rebound curves. When the apparatus RH matched ambient RH, 95 % of the particles adhered as a campaign average. Secondary organic material, produced for the most part by the oxidation of volatile organic compounds emitted from the forest, produces liquid PM over this tropical forest. During periods of anthropogenic influence, by comparison, the rebound fraction dropped to as low as 60 % at 95 % RH. Analyses of the mass spectra of the atmospheric PM by positive-matrix factorization (PMF) and of concentrations of carbon monoxide, total particle number, and oxides of nitrogen were used to identify time periods affected by anthropogenic influences, including both urban pollution and biomass burning. The occurrence of nonliquid PM at high RH correlated with these indicators of anthropogenic influence. A linear model having as output the rebound fraction and as input the PMF factor loadings explained up to 70 % of the variance in the observed rebound fractions. Anthropogenic influences can contribute to the presence of nonliquid PM in the atmospheric particle population through the combined effects of molecular species that increase viscosity when internally mixed with background PM and increased concentrations of nonliquid anthropogenic particles in external mixtures of anthropogenic and biogenic PM.
Particulate matter (PM) directly affects the Earth's climate by scattering and absorbing solar radiation and indirectly through effects on clouds (Ramanathan et al., 2001). The magnitude of these effects depends in part on the physical and chemical properties of the particulate matter (Andreae and Rosenfeld, 2008). The physical state of PM, as liquid or nonliquid, can influence the growth rates of small particles and ultimately the production of cloud condensation nuclei (CCN) (Riipinen et al., 2011; Perraud et al., 2012). Liquid particles pose negligible in-particle diffusion barriers for condensing species and therefore can grow rapidly. By comparison, nonliquid particles, referring to both semisolid and solid particles, can have different behavior. For some conditions, semisolid particles can grow slowly because of in-particle limits on rates of molecular diffusion, and solid particles can grow even more slowly when limited to surface adsorption (Riipinen et al., 2012; Shiraiwa and Seinfeld, 2012; Li et al., 2015). Liquid compared to nonliquid PM can also affect reactivity (Kuwata and Martin, 2012; Li et al., 2015). As a consequence of the differing growth mechanisms, the growth of small particles can be relatively disfavored in a population of liquid particles of heterogeneous sizes compared to a similar population of nonliquid particles (Zaveri et al., 2014). An implication can be that CCN concentrations are ultimately greater for a population of nonliquid particles that grows to CCN sizes, as compared to a population of liquid particles.
Secondary organic material (SOM), produced by the oxidation of biogenic volatile organic compounds (BVOCs), is a major source of atmospheric PM, especially over forested regions where SOM often dominates the mass concentration of submicron PM (Hallquist et al., 2009; Jimenez et al., 2009). The physical state of SOM has been studied in both laboratory (Vaden et al., 2011; Kuwata and Martin, 2012; Perraud et al., 2012; Saukko et al., 2012; Renbaum-Wolff et al., 2013; Kidd et al., 2014; Bateman et al., 2015; Li et al., 2015; Liu et al., 2015; Pajunoja et al., 2015; Song et al., 2015) and ambient environments (Virtanen et al., 2010; O'Brien et al., 2014; Bateman et al., 2016; Pajunoja et al., 2016). For background conditions of the Amazonian tropical forest, a region dominated by isoprene-derived SOM and high RH, PM was mostly liquid (Bateman et al., 2016). For a boreal forest in northern Europe, a region dominated by pinene-derived SOM and low RH, PM was largely nonliquid (Virtanen et al., 2010). The combined set of laboratory and ambient studies shows that the physical state of PM with high SOM content depends on the surrounding relative humidity (RH). This effect arises in part because organic particles are hygroscopic to various extents depending on composition. Water absorption, which is favored at elevated RH, has a plasticizing effect on physical state (Koop et al., 2011).
The physical state of PM affected by urban pollution over forests remains largely unexplored. A single day of ambient observations in central Amazonia suggested that ambient PM affected by urban pollution tended toward a nonliquid state (Bateman et al., 2016), and laboratory studies support the idea of an important modulating role of pollution in the physical state of SOM. A nonliquid state was favored for SOM produced from single-ring aromatic species (Liu et al., 2015; Song et al., 2015) as well as SOM mixed with polycyclic aromatic hydrocarbons (PAHs) (Zelenyuk et al., 2012; Abramson et al., 2013; Berkemeier et al., 2014). Organic molecules associated with urban pollution and industrial activities tend to be less hygroscopic than biogenic SOM (Hersey et al., 2013). When internally mixed within biogenic SOM, the anthropogenic molecules have a tendency to reduce water uptake and thereby increase the viscosity of the mixed particles. A similar line of reasoning leads to an identical hypothesis for the effects of biomass burning emissions. Compared to prevailing background conditions of SOM dominance over many forests, PM produced by biomass burning leads to a net effect of decreased water uptake when the PM mixes into the background particle population (Dusek et al., 2011).
The data sets presented herein provide observational evidence on the effects
of anthropogenic influences on the physical state of ambient particulate
matter. All other factors being equal, the lack of particle rebound is an
indicator of liquid PM (Bateman et al., 2015). Conversely, the occurrence of
particle rebound is an indicator of nonliquid PM. Rebounding and adhering
particles were separately collected for conventional and chemical imaging.
The data sets were collected during the two intensive operating periods
(IOP1 and IOP2) of the GoAmazon2014/5 experiment, corresponding to the wet
and dry seasons (Martin et al., 2016). The research site
(
An impactor apparatus was used for the study of particle rebound (Bateman et
al., 2014, 2015, 2016; Li et al., 2016). The apparatus was housed inside a
temperature-controlled research trailer at the T3 site. Particulate matter
was sampled at 5 m above ground through copper tubing with an outer diameter
of 13 mm (0.5 in.). In sequence, a drying unit reduced the sampled flow to
25 % RH or lower, a differential mobility analyzer (DMA; TSI 3085)
selected a subpopulation of dried particles having a narrow distribution of
electric mobility, and a humidification unit (Nafion tubes; Perma Pure,
MD 110) elevated the RH of the mobility-filtered flow to the targeted RH of a
measurement. The drying unit consisted of a Nafion drier in series with a
silica gel diffusion drier, and the silica gel was replaced every 2 days.
After passing the dryer, DMA, and humidifier, the resulting flow was split
and passed through three impactors operated in parallel. Labels
The three impactors differed from one another by having uncoated, coated, or
no impaction plate. The impactor with the uncoated plate passed both
nonimpacted and rebounded particles. The impactor with the coated plate
(Dow Corning high-vacuum grease) passed only nonimpacting particles. The
impactor with no plate passed all particles. Its purpose was to serve as a
compensation arm for possible miscellaneous particle losses, such as wall
loss. Based on the particle number concentrations measured downstream of the
impactors, the rebound fraction was calculated as follows (Bateman et al.,
2014):
The terms
A rebound curve representing
In conjunction with the rebound measurements, particles were collected for imaging by transmission electron microscopy (TEM) and scanning transmission X-ray microscopy (STXM). Images in some cases can directly suggest the liquid or nonliquid state of individual particles (O'Brien et al., 2014; Wang et al., 2016). Samples for imaging were collected as follows. A fourth impactor was added in parallel with the other three impactors. Imaging substrates were affixed to the impaction plate, and multiple substrates were mounted to the plate for concurrent collection for the various microscopy techniques. Substrates included grids coated with formvar (EMJapan Co., Tokyo, Japan) or lacey carbon (Ted Pella Inc., USA) for TEM and silicon nitride membrane windows (Silson, UK) for STXM. The flow rate and set point aerodynamic diameter of the fourth impactor were the same as for the other three impactors. The collected particles represented those that adhered to the substrate at impact, and an assumption in the analysis is that rebound from the substrate was similar to that from the uncoated plate. The flow through this fourth impactor was pulled by an in-line TEM autosampler (Arios Inc., Tokyo, Japan; Adachi et al., 2014). In this way, the setup separately collected particles that adhered to the grids on the impaction plate and particles that rebounded from the impaction plate and passed to the autosampler. The autosampler collected particles having aerodynamic diameters from 60 to 350 nm. Particles adhered to the TEM substrates in the autosampler, yet rebounded from the TEM substrates in the impactor because of the significantly different particle impact velocities between the two impactors (i.e., cut point and impactor design). Samples were collected for TEM analysis between 30 September and 15 October 2014. Samples were collected for STXM analysis between 01:00 and 10:00 UTC on 1 October 2014.
Microanalysis of individual particles of the collected PM was performed using two instruments: (1) a transmission electron microscope (TEM; JEOL, JEM-1400) equipped with an energy-dispersive X-ray spectrometer (EDS; Oxford Instruments) and (2) a scanning transmission X-ray microscope interfaced for near-edge X-ray absorption fine structure spectroscopy (STXM/NEXAFS; Advanced Light Source, Berkeley). For the TEM, imaging was conducted through bright-field microscopy, and particle composition was investigated by EDS. For the STXM at a fixed photon energy, an image was obtained by detecting the transmitted light at each pixel while raster scanning the sample. Spatially resolved NEXAFS spectra were obtained from a set of images recorded at different photon energies. The NEXAFS spectra provided chemical bonding information and quantitative elemental ratios (Moffet et al., 2010a, b; O'Brien et al., 2014; Piens et al., 2016). Section S2 of the Supplement presents further technical information concerning the TEM and STXM/NEXAFS analyses.
Additional colocated measurements used in the data analysis herein included a
high-resolution time-of-flight aerosol mass spectrometer (AMS; Aerodyne
Inc.), a single-particle soot photometer (SP2; Droplet Measurement
Technologies), a size-resolved cloud condensation nuclei counter (CCNC;
Droplet Measurement Technologies, CCNC-100), a condensation particle counter
(TSI, CPC 3772) for measuring particle number concentrations, an integrated
cavity output spectroscope (ICOS; Los Gatos) for measuring carbon monoxide
(CO) concentration, and a trace-level enhanced detector (TLE; Thermo
Scientific, Model 42i, with further customization) for the measuring
concentrations of nitrogen oxides (NO
Particle rebound or particle adhesion at impact depends on the balance of
energies. Particle rebound occurs when the kinetic energy before impact is
greater than the sum of dissipation and surface adhesion energies after
impact (Tsai et al., 1990; Bateman et al., 2014). The dissipation energy of
liquid particles is much greater than that of solid particles because of
additional mechanisms of dissipation available to the former. As a point of
reference, the transition from rebound to adhesion occurs across a viscosity
transition of 10
The rebound curves of particles with 190 nm in mobility diameter are shown in
Fig. 1a for IOP1 and IOP2. For RH < 50 %, the rebound fraction
was between 0.8 and 1.0. For RH > 50 %, the rebound fraction
decreased monotonically to a low value, typically 0. The shape of the
rebound curve in Fig. 1a is similar to that for particles of secondary
organic material produced in the Harvard Environmental Chamber (Bateman et
al., 2015). The rebound fraction of SOM particles produced from isoprene or
Rebound fraction as a function of apparatus relative humidity during
IOP1 (blue) and IOP2 (red) corresponding to the wet and dry seasons,
respectively, in central Amazonia. Panel
The subset of the data for which the apparatus RH matched the ambient RH is shown in Fig. 1b (see Sect. S3 of the Supplement). In light of the distribution of ambient RH values (see Fig. 1c), the data set in Fig. 1b implies that submicron PM in this tropical environment was liquid most of the time. Bateman et al. (2016) reported such a result for observations under background conditions at this site for data sets collected in 2013 a few months before GoAmazon2014/5 began. The results reported for GoAmazon2014/5 represent a longer time series (i.e., 550 compared to 30 rebound curves) and reinforce the generality of the earlier results found by Bateman et al. (2016) on the prevalence of liquid PM for this forested region under background conditions.
In the larger set of observations in the present study, what also emerges is that the rebound fraction remained at a median value of 0.05, even at 95 % RH, implicating the presence of externally mixed PM in the atmosphere. Approximately 5 % of the particles were nonliquid, even to 95 % RH, in both the wet and dry seasons (i.e., IOP1 and IOP2) (Fig. 1a and b). At times, rebound fractions of up to 0.4 occurred for RH > 90 %. Elevated rebound fractions were observed more frequently in IOP2 than IOP1. During IOP2, which was extensively influenced by biomass burning, nonliquid particles at the median constituted a fraction of 0.3 at 70 % RH and a fraction of 0.1 at 90 %. These RH values prevailed approximately 10 % of the time. Events of elevated rebound fraction were associated with the pollution plume from Manaus during IOP1 in the wet season and with both this urban plume and increased regional biomass burning during IOP2 in the dry season (see Sect. 3.2).
Probability density functions (PDFs) of the rebound fraction at 75 % RH based
on the data sets of Fig. 1a are shown in Fig. 2 for (a) IOP1 and (b) IOP2 for
three types of air masses during day- and nighttime-periods. The daytime
period is 12:00 to 16:00 LT (local time) (16:00 to 20:00 UTC) and
the nighttime period is 23:00 to 04:00 LT (03:00 to 08:00 UTC). Air masses
were identified as influenced by local to regional biomass
burning, as influenced by Manaus pollution, or as background. The classification scheme was
based on concentration regimes of particle number, carbon monoxide, and odd
nitrogen (NO
The probability density function of rebound fraction during
The increase in rebound at night might be explained by a combination of interacting factors. A shallow stable nocturnal boundary layer can trap and thereby concentrate anthropogenic nonliquid particles from local emissions at night, including smoldering fires during IOP2. During the day, the boundary layer expands and dilution is more effective. In addition, the production of liquid secondary organic material by biogenic processes is comparatively more rapid.
For further analysis, the rebound curves recorded under background
conditions were averaged separately for IOP1 and IOP2 (see Fig. 1a and
Sect. S5 of the Supplement). These two background-average curves served as
references against which deviations in rebound fraction were calculated for
all air masses. Rebound deviation represents the excess nonliquid PM over
background conditions after detrending the data for dependence on
relative humidity. The rebound deviations for IOP1 and IOP2 are plotted in
Fig. 3. Rebound deviations as high as
Transitions between background and polluted conditions across 24 h periods are presented in Fig. 4 as representative examples for each IOP. The bottom panel shows the deviation in rebound fraction relative to the background-average curve (see Sect. S5 of the Supplement). Color coding distinguishes relative humidity. From bottom to top, other panels in the figure show temperature, wind direction, wind speed, relative fractions of two groups of AMS PMF factor loadings, black carbon (BC) concentration, sulfate concentration, and total submicron PM mass concentration. Processes contributing to the loading of PMF group A are largely associated with the background atmosphere of Amazonia, including the possibility of long-range transport and extensive oxidation of biomass burning emissions. Processes contributing to the loading of PMF group B are largely associated with urban pollution and local and regional biomass burning (see Sect. S6 of the Supplement).
Deviation in rebound fraction relative to the average curve for
background conditions during
Time series of representative species during two pollution events for IOP1 and IOP2 (left and right sides of figures, respectively). Total mass concentration plotted in the figure represents the sum of the AMS and black carbon mass concentrations measured for submicron PM. The PMF mass fractions are expressed in relative terms to one another and necessarily sum to unity. Time on the abscissa is expressed in UTC.
Shifts in the deviation of the rebound fraction from the background-average curve are apparent in the time series in the bottom panels of Fig. 4. At these times, the fractional loading of PMF group A decreased and that of PMF group B increased, indicating a shift away from background conditions. Background conditions were characterized by high loadings of PMF group A and small rebound deviations. In the left panel (IOP1) is an example of one type of anthropogenic event: black carbon concentration and the fractional loading of PMF group B abruptly increased together. The rebound deviation simultaneously increased, indicating an increasing presence of nonliquid PM, especially above 70 % RH. In the right panel (IOP2) is an example of a second type of anthropogenic event: the background air mass was gently replaced by an air mass characterized by small increases in rebound deviation and fractional loading of PMF group B, yet it was lacking an associated increase in black carbon concentration. In an example of a third type of anthropogenic event, this air mass was later replaced by an air mass characterized by a strong increase in the fractional loading of PMF group B, total PM mass concentration, and rebound deviation. Around 15:00 UTC, a convective event associated with rainfall decreased temperature, changed wind direction, and increased wind velocity; background conditions returned.
TEM images collected during the time periods of elevated rebound corroborate
the foregoing interpretation of liquid and nonliquid PM as adhering
compared to rebounding particles. Images of particles that adhered to the
TEM substrates in the impactor at 95 % RH are shown in Fig. 5a. The
images were taken at a tilt angle of 60
Representative TEM images of
Deviation in rebound fraction for categorization by type of air mass and time of day. The box–whisker representation of the 10, 25, 50, 75, and 90 % quantiles of statistics for each RH bin is explained in the caption to Fig. 1. Air mass categorization is as for Fig. 2. Results are shown for particles having a mobility diameter of 190 nm.
The results of Figs. 3 and 4 are further analyzed in Fig. 6. Rebound deviation statistics are shown by box–whisker representations for different windows of relative humidity. Figure S1 of the Supplement presents an additional level of detail. The data sets were segregated for presentation by IOP1-IOP2, daytime-nighttime, and air mass type. For background conditions, the rebound deviations relative to their average were mostly 0, indicating that there was low variability among different background air masses. The exception was for the night periods of IOP1. By comparison, under anthropogenic influences, the rebound deviation was positive for both IOPs. Positive deviations were most significant between 65 and 95 % RH. In all cases, the nighttime deviations were greater than the daytime counterparts. For IOP2, the prevalence of biomass burning confounded separate classifications of urban pollution and biomass burning, and a classification of biomass burning took precedent. Rebound deviations were strongest during these time periods. Statistics of the analysis are further summarized in Table S2 (see Sect. S7 of the Supplement). The average O : C ratios for the various air-mass classifications can be found in Table S3 of the Supplement.
Scatter plots of rebound deviation with environmental variables of ambient temperature, wind speed, and wind direction show no correlation for daytime or nighttime data sets (see Fig. S2 of the Supplement). Scatter plots of rebound deviation with some possible anthropogenic influences are presented in Fig. S3. Soot, typically characterized by a solid core region of black carbon, is expected in abundance both in urban pollution and biomass burning emissions. There was, however, no correlation between rebound deviation and black carbon concentrations. There was also no correlation between rebound deviation and total particle mass concentration. Rebound deviation and sulfate concentration weakly anticorrelated, which might be expected given the hygroscopicity of sulfate. Sulfate concentrations, however, were a poor indicator of anthropogenic influence in this region because the variability in background concentrations was comparable in magnitude to any urban influence (de Sá et al., 2016). As a caveat, an assumption in correlation tests is that variance arises from a single variable, and the possibility of two or more contributing or interacting factors is not directly considered.
Given that water uptake is an important process for softening organic
material, scatter plots of the rebound deviation at 75 % RH with the
hygroscopicity parameter
The apparent shift to higher
An analysis of the relationship between rebound deviation and chemical characteristics is presented in Fig. 8 based on the fractional loading of PMF group B. The data sets are segregated for presentation by IOP1-IOP2, daytime-nighttime, and four bands of fractional loading. Within each panel, box–whisker statistics of rebound deviation are shown for different windows of relative humidity, ranging from 50 to 95 %. The figure shows that the rebound deviation increased at all RH values as the fractional loading of group B increased. The background-average curve used as the reference for rebound deviation corresponded to a fractional loading of 0.00 to 0.15 for group B and 0.85 to 1.00 for group A. An increasing fractional loading of group B represented greater anthropogenic influence. The inference is that anthropogenic influences, represented by a combination of urban pollution and biomass burning, affected chemical composition in ways that increased the presence of nonliquid PM above 50 % RH.
Scatter plot of rebound deviation with the hygroscopicity parameter
Deviation in rebound fraction for categorization by chemical characteristics. The box–whisker representation of the 10, 25, 50, 75, and 90 % quantiles of rebound fraction is explained in the caption to Fig. 1. The chemical characteristics are categorized by the fractional loading of PMF group B as < 0.15 (green), 0.15 to 0.3 (blue), 0.3 to 0.6 (orange), and > 0.6 (red).
Scatter plots between rebound deviation at 75 % RH and the fractional
loading of group B are shown in Fig. 9a and b for the data sets of the two
IOPs. The data points are colored according to the value of the
hygroscopicity parameter
Scatter plot of rebound deviation with the fractional loading of PMF
group B for
A model to predict rebound deviation based on chemical characteristics was
constructed. The fractional loadings of PMF group A and B were used as model
inputs, and model coefficients represented the effects of RH across nine
bands (see Sect. S9 of the Supplement). The observed and predicted rebound
deviations are plotted in Fig. 10. The corresponding coefficients
Scatter plot of observed compared to predicted rebound deviation for
Analysis by STXM/NEXAFS supports the foregoing narrative of anthropogenic influence as a modulator between liquid and nonliquid PM. Rebounded particles were collected on 1 October 2014 during a time period classified as influenced by biomass burning emissions. The carbon K-edge spectrum is shown in Fig. 11a, and the STXM image is shown in Fig. 11b. A notable feature of the NEXAFS spectrum of the rebounded particles is the strong double bond. Pöhlker et al. (2012) previously collected NEXAFS spectra for samples collected at a background site in central Amazonia, and the strong feature of a double bond was absent. The spectra instead resembled those of different types of reference biogenic secondary organic material.
STXM/NEXAFS analysis of particles collected after rebound from the
impaction plate. Samples were collected between 01:00 and 10:00 UTC on
1 October 2014.
For comparison to the spectrum collected of the rebounded particles, carbon
K-edge spectra are shown for carbonaceous particles collected in other field
and laboratory studies. Based on these results as well as those of the
rebound measurements, a hypothesis of soot or black carbon to explain the
rebounding particles was ruled out on three grounds: the double-bond feature
was homogeneously distributed throughout the particles (Fig. 11b) compared to
inclusions that are typical for soot (Moffet et al., 2013; Knopf et al.,
2014; O'Brien et al., 2014), rebound deviation and black carbon
concentrations did not correlate (Fig. S3), and the spectroscopic signatures
of rebounded particles and soot did not match (Fig. 11a). A hypothesis of
VOC-derived secondary organic material, including possible changes because of
shifts from HO
There was exceptional uniformity in the particle population characterized by
STXM/NEXAFS, which could suggest that the rebounding PM represented distant
sources or, alternatively, a strong single nearby source. The gray region
around the red line in Fig. 11a illustrates the low variability across the
population of analyzed particles (Fig. 11b). Moreover, the variability in
the O : C ratio determined by the NEXAFS analysis was just
Several speculations can be made for the origins of the particles leading to the data set from 1 October. The chemical constituents giving rise to the double bonds might derive from biological degradation products from incomplete combustion, such as in biomass burning (Tivanski et al., 2007; Keiluweit et al., 2010). Unexplained by this speculation, however, is the absence of a potassium signature in the NEXAFS spectra, which is typical of most biomass burning. Future collection of NEXAFS spectra would be advisable for the several different types of biomass burning in an Amazonian context, such as from nearby fields, regionally around Manaus, from other regions of South America 2 or 3 days away, and from Africa up to a week away. An alternative speculation for this data set is that solid organic particles produced by the impact of raindrops on wet soil surfaces could be making a contribution to the rebounded PM analyzed here (Joung and Buie, 2015). Wang et al. (2016) recently reported the detection of airborne soil organic particles generated by this mechanism over agricultural fields in the central plains of the USA, and the corridor from Manaus to T3 has many agricultural fields. The rebounded particles collected at T3 and the agricultural particles reported in Wang et al. (2016) both had a homogeneous distribution of double bonds and similar elemental ratios and absorption features. Even so, preliminary analysis across the extended data set at T3 between rebound and nearby precipitation did not show a clear correlation. Another speculation relates to the importance of aromatic compounds as hardening agents. Several gas-phase aromatic compounds, laden with double bonds, were measured during IOP1 and IOP2, including toluene, benzene, trimethylbenzene, and xylenes, by proton-transfer mass spectrometry (Y. Liu et al., 2016). Rebound deviation correlated positively with the concentrations of these compounds during both IOPs, and the correlation was stronger during the night. Laboratory studies show that the uptake of polycyclic aromatic hydrocarbons during the formation of biogenic PM can increase the viscosity of the PM (Vaden et al., 2011; Zelenyuk et al., 2012; Abramson et al., 2013; Liu et al., 2015).
Under background conditions, particles composed primarily of highly oxidized
biogenic PM were hygroscopic, and they were liquid for the RH values
prevailing over Amazonia at surface level. Anthropogenic influences of urban
pollution and biomass burning decreased hygroscopicity, and nonliquid PM
became more favored. The shift in physical state correlated with increasing
concentrations of C
The underlying time series of data is provided as text files in the Supplement.
The authors declare that they have no conflict of interest.
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). The Office of Biological and Environmental Research of the Office of Science of the United States Department of Energy is acknowledged for funding, specifically the Atmospheric Radiation Measurement (ARM) Climate Research Facility, the Atmospheric System Research (ASR) Program, the Division of Chemical Sciences, Geosciences, and Biosciences (Advanced Light Source at Lawrence Berkeley National Laboratory, Beamlines 5.3.2 and 11.0.2), the Environmental Molecular Sciences Laboratory (EMSL), and Pacific Northwest National Laboratory (PNNL). Further funding was provided by the Amazonas State Research Foundation (FAPEAM), the São Paulo State Research Foundation (FAPESP), the Brazil Scientific Mobility Program (CsF/CAPES), the US National Science Foundation, and the Japanese Ministry of the Environment. The work was conducted under scientific licenses 001030/2012-4 of the Brazilian National Council for Scientific and Technological Development (CNPq). Edited by: T. Karl Reviewed by: two anonymous referees