Particle morphology is an important parameter affecting aerosol
optical properties that are relevant to climate and air quality, yet it is
poorly constrained due to sparse in situ measurements. Biomass burning is a
large source of aerosol that generates particles with different morphologies.
Quantifying the optical contributions of non-spherical aerosol populations is
critical for accurate radiative transfer models, and for correctly
interpreting remote sensing data. We deployed a laser imaging nephelometer at
the Missoula Fire Sciences Laboratory to sample biomass burning aerosol from
controlled fires during the FIREX intensive laboratory study. The laser
imaging nephelometer measures the unpolarized scattering phase function of an
aerosol ensemble using diode lasers at 375 and 405 nm. Scattered light from
the bulk aerosol in the instrument is imaged onto a charge-coupled device
(CCD) using a wide-angle
field-of-view lens, which allows for measurements at 4–175
Atmospheric aerosol particles absorb and scatter light, strongly influencing
Earth's climate (Davidson et al., 2005; Stocker et al., 2013). Size,
composition, and morphology are key parameters in determining aerosol optical
properties. Aerosol morphology is particularly important in implementing
global radiative transfer calculations, since the directionality of scattered
light from aerosol relative to the solar zenith angle will vary with particle
shape (Kahnert and Devasthale, 2011). Most satellite and ground-based remote
sensing retrievals assume that accumulation-mode aerosol
(
Biomass burning is a major source of aerosol, with average annual global emissions of 49 Tg (Stocker et al., 2013). Biomass burning particle emissions are composed predominantly of a mixture of fractal-like black carbon and spherical organic aerosol, along with a minor contribution from other species such as dust and inorganic salts (Adachi and Buseck, 2011; China et al., 2013; Hand et al., 2005; Li et al., 2003; Pósfai et al., 2004). Biomass burning and biofuel combustion accounts for 67 % more black carbon particle emissions than fossil fuel combustion, and is also a major source of organic aerosol (Bond et al., 2004; Liousse et al., 1996). The complex morphology of biomass burning aerosol and its high atmospheric abundance make it an important target for improved characterization of optical properties.
In the past, morphology was generally inferred from dynamic shape factor experiments based on, for example, aerodynamic size and mass measurements (Hand and Kreidenweis, 2002; Slowik et al., 2004; Zelenyuk et al., 2006). These methods require multiple instruments, and are not well suited for measurements of evolving samples where high time resolution is desirable. Furthermore, the dynamic shape factor provides some information about the effective density of the particle, but does not provide important information on the optical properties of the particles. Alternatively, samples have been analyzed offline using scanning or transmission electron microscopy (SEM/TEM; Brodowski et al., 2005; McDonald and Biswas, 2004). Quantitative characterization using these methods relies on sophisticated software, can be quite time-consuming, and assumes that the morphology is not altered by deposition to the imaging substrate. Lidar can be used to derive a scalar nonsphericity value based on the depolarization ratio of backscattered radiation. This value is used in remote sensing to indicate, for example, when mineral dust aerosols are present (Mishchenko and Sassen, 1998; Omar et al., 2009), but does not work well for accumulation-mode particles (Murayama et al., 2003). Recently, a sophisticated method for determining the morphology of coarse-mode aerosol using holographic imaging without interrupting the aerosol flow was demonstrated (Berg et al., 2017); however this method is unfeasible for accumulation-mode particles.
The aerosol phase function is the angular distribution of the radiation
scattered from the particles. It is dependent on the size distribution,
complex refractive index, and morphology of the aerosol population.
Instruments used for measurements of aerosol scattering phase function can be
broadly grouped into four categories. The most common technique uses
hemispheric measurements of backscatter and total scatter from integrating
nephelometers (Anderson et al., 1996). Some nephelometers also have an
option to measure set angular regions by blocking portions of the forward
scattered light (Chamberlain-Ward and Sharp, 2011). While these devices are
reliable and robust, the measurements over large portions of the phase
function do not provide much information about particle morphology due to the
lack of angular specificity. Second, some studies use moveable detectors
along a ring structure such that the phase function can be measured by
sweeping the detector (or multiple detectors) across a broad range of angles
(Holland and Gagne, 1970; Hovenier et al., 2003; Jaggard et
al., 1981; Kuik et al., 1991; Perry et al., 1978; Volten et al., 2001). While
this allows for good angular coverage (up to 3–173
Laser imaging nephelometer instrument diagram showing beam paths of the 375 and 405 nm beams, which are cycled in sequence in the experiment.
Here we describe a modified commercial laser imaging nephelometer based on the same principles as the polarized imaging nephelometer that measures the unpolarized phase function at 375 and 405 nm. This method has the advantages of being online, realtime, and sufficiently sensitive for ambient aerosol loadings. We use this instrument to determine morphology characteristics of accumulation-mode aerosol by imaging the aerosol phase function and directly measuring the effects of morphology on radiative properties. This allows for a clearer understanding of how the dominant morphology of the measured aerosol sample affects, for example, radiative transfer calculations.
We deployed the laser imaging nephelometer as part of the Fire Influence on Regional and Global Environments Experiment (FIREX) intensive study at the Missoula Fire Sciences Laboratory to measure biomass burning aerosol. We compare the retrieved phase functions to models of spherical and fractal particles to provide direct verification that the predominant morphology of the freshly emitted particles varied by fuel type, and demonstrate the ability to assign particle morphology from the phase function.
The laser imaging nephelometer (LiNeph; AirPhoton, Baltimore, MD, USA), shown
schematically in Fig. 1, measured nearly the entire unpolarized scattering
phase function (
Example raw CCD image during Fire A for the 405 nm laser. The
yellow rectangle indicates the direction across which the image is integrated
in order to determine scattering for a single angle bin
(
A wide angle field-of-view fish-eye lens (FE185C046HA-1; Fujifilm, Tokyo,
Japan) collected the light and imaged it onto a 16-bit, 2750
The raw CCD images were processed in five steps to produce an aerosol
scattering phase function by correcting for the CCD dark background,
scattered light due to internal surfaces, and Rayleigh scattering. First, a
dark background spectrum was subtracted from each CCD image. This is
necessary because CCD detectors produce non-zero dark current in the absence
of light due to thermal noise and an electronic offset. The dark background
image was acquired with the same integration time as the aerosol images and
was measured before each fire. Second, light scattering from surfaces within
the sample volume was removed by subtracting a laser power-normalized
correction from each image. The scattering correction was determined by
filling the chamber with helium, which has a negligible scattering
coefficient, and taking an average of
The wavelength-dependent pixel-to-angle calibration was measured in the
laboratory before and after the FIREX measurements using standard
monodisperse polystyrene latex spheres (PSLs) with diameters ranging from 100
to 895 nm. For these experiments, a solution of monodisperse NIST-traceable
polystyrene latex (PSL; 3000 Series Nanospheres, Thermo Fisher Scientific, Waltham, MA, USA) in
pure water was atomized, dried, and size-selected by a differential mobility
analyzer (DMA) before being sampled by the imaging nephelometer and a
condensation particle counter (CPC 3022; TSI Inc., Shoreview, MN, USA). The
measured phase functions were compared to modeled phase functions from Mie
theory for each monodisperse samples of PSLs used to determine a
wavelength-dependent angular calibration. There was a small difference in the
calibrations at 375 and 405 nm due to achromatic behavior of the wide-angle
lens. The pixel-to-angle relationship was found to be linear (see
Supplement), with each pixel bin corresponding to
The total integrated aerosol scatter was determined for each image by integrating the illuminated area and applying a size-dependent truncation using the measured aerosol size distribution and assuming spherical particles. While the total scattering measured at 375 nm is greater, the overall shape of the phase functions at 375 and 405 nm are quite similar within the signal level of the measurements. Therefore only phase functions measured at 405 nm will be shown, and we focus on the analysis of the 405 nm phase functions here.
The Missoula Fire Sciences Laboratory is a United States Department of
Agriculture (USDA) facility with a large (
Schematic of NOAA aerosol sampling inlet at Missoula Fire Sciences Laboratory.
Biomass burning aerosol was sampled from the smoke-filled room through
A collection of aerosol instruments sampled from the mixing volume, including the imaging nephelometer, an integrating nephelometer, and cavity ring-down photoacoustic spectrometer (CRD PAS; Lack et al., 2012; Langridge et al., 2011). A laser aerosol spectrometer (LAS; TSI Inc., Shoreview, MN, USA; Jonsson et al., 1995) periodically measured the aerosol size distribution from the mixing volume during each fire. Likewise, a single particle soot photometer (SP2; Droplet Measurement Technologies, Longmont, CO, USA) measured the concentration of refractory black carbon aerosol, and provided an estimate of the enhancement of light absorption due to lensing by internally mixed material coating refractory black carbon cores (Schwarz et al., 2006, 2008).
The laser imaging nephelometer sampled in series with a traditional commercial integrating nephelometer (3563; TSI Inc, Shoreview, MN, USA). This instrument, which is commonly used in laboratory and field studies, contains a halogen lamp to illuminate the sample and measures the total scatter and backscatter at red (750 nm), green (550 nm), and blue (450 nm) wavelengths using three different photomultiplier tubes with narrow bandpass filters (Anderson and Ogren, 1998; Heintzenberg and Charlson, 1996). The flow through the nephelometers was 10–15 slpm, controlled by a mass flow controller (Alicat Scientific, Tucson, AZ, USA) and diaphragm pump (Gast Manufacturing, Benton Harbor, MI, USA). The integrating nephelometer was upstream of the imaging nephelometer and a baratron (MKS, Andover, MA, USA) recorded the pressure at the outlet of the imaging nephelometer.
During some fires, aerosol particles were collected on silicon wafers for
offline analysis using a scanning electron microscope (SEM). These were
collected on a separate line from the general inlet with a 1
Biomass burning aerosols consist predominantly of a mixture of primary fractal-like black carbon particles and spherical organics formed by condensation of semi-volatile species from the gas phase (Adachi and Buseck, 2011; Hand et al., 2005; Li et al., 2003; Pósfai et al., 2004). In this paper, we adopt the definition of black carbon provided by Bond et al. (2013): “a distinct type of carbonaceous material that is formed primarily in flames, is directly emitted to the atmosphere, and has a unique combination of physical properties”. To look at how morphology affects the retrieved phase function, we used two morphology parameterizations: Mie theory for spherical particles and Rayleigh–Debye–Gans (RDG) for fractal-like particles. The spherical particles are assumed to be predominantly composed of organic matter, whereas black carbon particles formed by incomplete combustion are fractal-like. Since fresh (< 4 h) emissions were measured in the absence of sunlight, it is unlikely that significant oxidative aging occurred in the smoke chamber. The smoke chamber relative humidity was less than 40 % for all fires measured, so we assume there was no significant restructuring of the black carbon agglomerates by organic or sulfuric acids (Xue et al., 2009; Zhang et al., 2008).
Mie theory describes electromagnetic radiation scattering by spherical
particles with size parameters (
Fractal clusters are defined by the following parameterization (Forrest and
Witten, 1979; Sorensen, 2001):
While fractal-like particles produced from fossil fuel combustion have been
studied extensively, few systematic measurements of the fractal
parameterization of fresh biomass burning combustion products have been
undertaken. Chakrabarty et al. (2006) reported
We used measured size distributions from the LAS to predict phase functions
for spherical and fractal-like biomass burning particles and compared those
with measurements by the imaging nephelometer. The LAS measures particle size
from the amount of light scattered by a 633 nm laser beam. The LAS size bins
had been calibrated using nominally spherical ammonium sulfate particles
(
The precision of the imaging nephelometer was quantified using an Allan–Werle
variance analysis (see Fig. S5) to the integrated signal
for individual angle bins (Allan, 1966; Werle et al., 1993). For
particle-free air, the standard deviation at 5
The two main sources of uncertainty for aerosol measurements are variations
in the number density and size of particles in the laser beam, and background
scatter attributed to temperature variability in the CCD. Errors arising from
fluctuations in the particles, notably for large particles, crossing the beam
can be reduced by averaging the signal over long periods. To compare with the
uncertainty for gas phase measurements above, the standard deviation with 1 min
averaging for a sample of 600 nm PSL at a concentration of 23 cm
The laser imaging nephelometer sampled aerosol from 11 separate fires using
the experimental configuration described above. In this analysis we focus on
two representative fires: pine (Fire A) and dry sagebrush (Fire B). These
fires were chosen qualitatively to juxtapose phase function measurements for
two classes of fuel type. For Fire A (#086), the fuel source was a mixture
of dead wood (logs) of different diameters, litter, duff, and canopy branches
with needles from lodgepole pine (
Total integrated scattering for
The accuracy of the total scattering measured by the imaging nephelometer is demonstrated by comparison with two independent measurements of aerosol scattering that were taken simultaneously. Figure 4a shows the total scatter measured by the imaging nephelometer at 405 nm compared with the integrating nephelometer and CRD PAS for Fire A. The scattering Ångström exponent (SAE) measured between 450 and 550 nm by the integrating nephelometer was used to extrapolate scatter at 405 nm for this instrument. The total scattering measured by the CRD PAS was determined by subtracting the photoacoustic absorption value, measured at 401 nm, from the cavity ring-down extinction at 405 nm. In all cases, measurements have been corrected for dilution in the shared inlet, and the retrieved scattering represents that of the original smoke sample. The good agreement between all three measurements provide confidence in the total scatter retrieved by the imaging nephelometer (see Fig. S1 for correlation plot).
Selected measured values demonstrating the key differences between the aerosol products from Fire A and Fire B. Note: TD stands for thermodenuded.
The square-wave appearance of the scatter during the experiment is a result
of alternating between the thermodenuder and unheated bypass channels on the
shared inlet. The large decrease in total signal when the aerosol was denuded
(
LAS volume size distribution for
In contrast, Fig. 4b shows that Fire B produced a more significant fraction of non-volatile aerosol indicated by the total scattering signal after the thermodenuder being only about 60 % less than after the bypass channel. The CRD PAS and imaging nephelometer agree well throughout this experiment. The higher scattering retrieved by the integrating nephelometer is likely due to errors arising from extrapolating from 450 nm. The integrating nephelometer data end at 13:58 MST because of an automatic recalibration. Interestingly, for Fire B the volume size distribution as measured by the LAS is much narrower and does not appear to shift significantly for the denuded products as was seen in the previous example. This could arise if the aerosol population has a predominantly fractal character. The side scattering at the angles measured by the LAS is much less size-dependent in the RDG model compared to Mie calculations, and therefore the LAS distribution would be expected to narrow for fractal-like particles. The SP2 absorption enhancement estimate is significantly lower for this fire (1.1–1.2), indicating very little coating on the refractory black carbon particles. In addition to the high fraction of scatter from the nonvolatile components, the AAE for this fire is much closer to unity, indicating absorption proportional to light frequency such as occurs for black carbon (see Table 1; Bergstrom et al., 2002).
Figure 6 shows measured phase functions at 405 nm averaged over one
individual cycle of the thermodenuder and bypass channel for well-mixed smoke
from Fire A. The measurements are overlaid with the model predictions
assuming spherical particles representative of organic aerosol (Mie theory)
and fractal-like black carbon particles (RDG theory) based on the LAS
particle size distribution. For aerosol measured after the bypass channel
(Fig. 6a), the Mie theory curve matches the measurement reasonably well.
Since the dominant aerosol types are highly volatile (i.e., evaporated at
Comparison of measured (blue) phase function at 405 nm to Mie
theory model (red) and RDG model (black) for one cycle of sampling through
bypass channel
This implication from the phase function measurements is further supported by
SEM images (Fig. 7), which show that the non-volatile (denuded) components
are mostly fractal-like particles. Using the SEM software, the average
monomer diameter was measured for monomers clearly visible on the outer edges
of the agglomerates; the average monomer diameter (excluding Pt coating) is
50
SEM image of aerosol collected after the thermodenuder during Fire A. While most of the particles appeared fractal-like, some spherical particles (presumably non-volatile organics) were also evident.
Figure 8 shows phase function plots for Fire B, which, in contrast to Fire A,
exhibits a significantly higher fraction of nonvolatile aerosol. In both
cases, the measured phase function is not described well using Mie theory,
exhibiting significantly higher scatter at angles < 30
Comparison of measured (blue) phase function at 405 nm to Mie
theory model (red) and RDG models (black and orange) for one cycle of
sampling through bypass channel
While the present study focused solely on fresh biomass burning emissions, it highlights that assuming similar optical properties (Mie theory) for all biomass burning aerosol may be inappropriate. Many remote sensing platforms assume spherical morphology for biomass burning aerosol (Dubovik et al., 2000; Omar et al., 2009; Remer et al., 1998). This study indicates that this is appropriate for the aerosol with high volatile organic content, but leads to highly inaccurate assumptions of the phase function for the dry brush fuel example, which is likely to have a higher contribution from black carbon. Since this work is primarily intended to demonstrate the capabilities of the instrument, we refrain at this point from drawing definite conclusions about atmospheric biomass burning aerosol from these two examples from controlled burns. It would be useful to collect more information about how the morphology of fresh emissions evolves with aging in the atmosphere. Studies have shown that biomass burning aerosols are coated with organic material within h in the atmosphere (e.g., Akagi et al., 2012; Vakkari et al., 2014). However, if there are conditions under which fractal-like particles do not immediately collapse or accumulate sufficient organic coatings to become spherical, then remote sensing retrievals of wildfire plumes from dry brush or grasses may significantly underestimate the forward scatter from the aerosol. Cheng et al. (2013) showed recently that assuming spherical morphology for simulated fractal-like soot particles leads to a significant underestimation in the aerosol absorption, as well as a large error in the scattering phase function, for remote sensing platforms that use reflectance. This error would likely have been unimportant for past retrievals from MODIS, for example, since cloud masking algorithms often misclassified thick smoke (typically fresh plumes) as clouds (Giglio et al., 2016). However, with improved biomass burning cloud masking algorithms, it will be interesting to see if MODIS retrievals of thick, fresh smoke plumes will be accurate with a spherical morphology algorithm. This will affect not only the accuracy of size distribution and total loading retrievals, but may also be important for radiative transfer models. Due to the more significant forward scatter component for fractal particle scattering phase functions, it is possible that ground-based scanning radiometer retrievals, such as from the Aerosol Robotic Network (AERONET), overestimate the contribution from coarse-mode particles in order to account for the high forward scatter measured using Mie theory. This biasing of the size distribution of biomass burning aerosol would also have important repercussions for estimates of radiative forcing.
This study demonstrates that a new laser imaging nephelometer can provide realtime, online phase function measurements of an accumulation-mode aerosol sample. The total scattering values retrieved compare well with both a traditional integrating nephelometer and with an extinction-minus-absorption method. In addition to total scatter, the ability to directly measure scattering phase function provides further information about the optical properties of the bulk aerosol that can be used to determine dominant particle morphology. Previously, particle morphology was inferred using a combination of aerodynamic and mobility measurements of particle diameter, or was measured offline using transmission or scanning electron microscopy. This new instrument offers a more convenient, quantitative method to determine the radiative effects of varying morphologies. While we demonstrated here the potential inaccuracy of assuming spherical scattering behavior for all biomass burning aerosol, this method also has potential for direct retrievals of phase functions for other non-spherical particles, such as mineral dusts, pollens, and ice crystals. Direct, accurate, in situ measurements of bulk aerosol phase functions offer new, important ways of improving the accuracy of both remote sensing retrievals and radiative transfer calculations.
Data are available at
The supplement related to this article is available online at:
The authors declare that they have no conflict of interest.
This project was supported by NOAA and the NASA Radiation Sciences Program. The authors would like to thank Karl Froyd, Chuck Brock, and Bernie Mason for useful discussions, as well as Dan Cziczo for lending us the Mini-MOUDI impactor. We are grateful to Jim Roberts and Carsten Warneke for organizing the FIREX intensive campaign in Missoula. Edited by: Manvendra K. Dubey Reviewed by: two anonymous referees