2010: CO & WA: Part 2 of 4: Scientific Paper: Measured and modeled humidification factors from biomass burning: role of inorganic constituents


2 Experimental methods

Controlled burns and measurements were conducted at the United States Forest Service

Rocky Mountain Research Station Fire Sciences Laboratory in Missoula MT from

21 May to 9 June 2006. Details describing the facility and experimental protocols have

10 been reported elsewhere (e.g. Day et al., 2006; Chen et al., 2006; Chakrabarty et

al., 2006; McMeeking et al., 2009; Petters et al., 2009; Lewis et al., 2009; Carrico et

al., 2010; Levin et al., 2010). Aerosol and gas-phase measurements were performed

after the chamber was filled with smoke and instruments sampled smoke directly. A

measured quantity of biomass (approximately 200 g) was burned on a continuously


weighed platform in the middle of the chamber. Fuels were ignited with a propane

torch at the edge of the fuel, resulting in a flame front propagating through the fuel and

both flaming and smoldering conditions occurring simultaneously. The fire was allowed

to extinguish naturally and the measurements continued for approximately two hours.

Typically three to four burns were performed a day. In this manuscript, we present


results from thirteen burns.

A thorough discussion of fuel types and their origin is provided by McMeeking et

al. (2009). We grouped the fuels into four main categories. The first category, “Forest/

Pine”, included ponderosa pine, lodgepole pine and southern pine. The second category,

“Brush”, included southern California chamise, southern California manzanita,


Utah juniper, Utah sage and rabbitbrush, and southern California ceanothus. “Southeastern/

Tropical”, the third category, included Puerto Rico fern and southeastern wax

myrtle. A final category of “Du

ff” included ponderosa pine duff and Alaskan duff. A


combination of woody material, needles and leaves were burned for each type, except


ff which included the decayed biomass on the forest floor as well as the top layer

of the soil. The fuel was dried before burning if it did not dry out su

fficiently during

shipping; fuel moisture contents were reported by McMeeking et al. (2009).


2.1 Light scattering coefficients

Two nephelometers (model M903, wavelength of 530 nm, Radiance Research Inc,

Seattle, Washington) measured light scattering coe

fficients. The experimental design

was similar to that used in a previous FLAME experiment (Day et al., 2006) and described

in detail by Day et al. (2000). Both nephelometers sampled through identical


plumbing that was connected to a PM2.5 cyclone (URG, Chapel Hill, North Carolina)

and a sampling inlet that allowed for the control of relative humidity. RH and temperature

were monitored at the entrance and exit of the nephelometers using hygroclip

sensors (Rotronic Instruments (UK) LTD, West Sussex, UK) with a reported accuracy


±1.5% RH at 23 C. The RH sensors were calibrated against a dew point hygrometer


plus RTD dry bulb temperature sensor (General Eastern Optica, Williston, Vermont) interfaced

to a humidity generator (Model 2000, Kaymont, Huntington Station, New York)

with reference probes calibrated using standard salt solutions. An error of

±3% is

estimated to reflect uncertainty in RH inside the nephelometer due to temperature fluctuations.

Also, RH at the inlet and outlet of the nephelometer can vary due to heating


in the nephelometer chamber. A small change in temperature (1 ) can result in a

significant change in RH at high RH (

5% at 95% RH) (Day et al., 2000). Although the

temperature and RH were monitored at the inlet and outlet of the nephelometer, the

exact RH and temperature inside the instrument may be slightly di

fferent. Therefore

we used the average RH of the inlet and outlet sensor to represent conditions in the


nephelometer chamber. The sample RH was controlled using diffusion tubes (Perma

Pure LLC, Toms River, New Jersey).

Calibrations of the nephelometers with dry, filtered air and SUVA (HFC 134a) span

gas were performed daily. The filtered air was used as a zero-point calibration and


SUVA was used as a high calibration point. During periods of low RH (

<25%) discrepancies


bsp values of 2–5% between the two nephelometers were observed (humidified

nephelometer was biased high) but were within expected uncertainties (Anderson

et al., 1996). Data from the humidified nephelometer were normalized to the dry


nephelometer data using linear regression equations derived from the comparisons of

data during dry conditions. For burns with no data corresponding to RH

<25% for both

instruments, corrections from an experiment earlier or later in the same day were applied.

The sample RH of the dry nephelometer measurement was maintained between

20–25%. After the biomass was ignited, the RH corresponding the humidified neph


elometer was increased from 20–25% to over 80% over a period of one to two hours

and the light scattering coe

fficients were measured as the aerosols were humidified.

Values of

bsp, RH and temperature were logged on a 5-s time interval.

Uncertainties in

bsp and f (RH) were computed by propagating errors derived from

calibration data following the procedure detailed in Day et al. (2006). Uncertainties in


bsp derived from calibration data (one standard deviation) were typically 5–8%. Comparisons

of the normalized data from ten burns from both nephelometers during periods

corresponding to low RH (

<25%) are shown in Fig. 1. Error bars reflect that the

measurement uncertainty accounted for di

fferences between the two instruments and

no additional biases in the data were observed. The uncertainty in

f (RH) was com20

puted by propagating the calibration and normalization uncertainties in

bsp and was


±0.08 or less for all burns. We estimated that background aerosols may

be contributing at most 0.03 to measured

f (RH) by assuming all of the bsp measured

during the chamber vents was due to ammonium sulfate. This contribution is likely an

overestimate and well within our experimental uncertainty.


2.2 Chemical composition

The IMPROVE (Interagency Monitoring of Protected Visual Environments) network

sampler (Malm et al., 2004) was used to collect smoke particulate matter for PM


chemical speciation and gravimetric mass analysis. The IMPROVE sampler consists


of four independent modules; the three used during FLAME were equipped with a

2.5 μm cyclone. Module A consisted of a Teflon filter that was analyzed for gravimetric

fine (PM

2.5) mass and elements with atomic number &#21;11 (Na) and &#20;82 (Pb) by XRF (Xray

florescence). Ion concentrations were determined using ion chromatography from


samples obtained from a nylon filter in module B. Module C utilized quartz fiber filters

for sample collection from which carbon was analyzed using thermal optical reflectance

(TOR) techniques to separate organic carbon (OC) from light-absorbing carbon (LAC)

(Chow et al., 2007). We assumed that the aerosols were internally mixed and composed

of inorganic salt species (KCl, K

2SO4, KNO3, (NH4)2SO4, NH4Cl and NaCl),


carbon (POM, particulate organic matter, and LAC) and soil (Al2O3 and CaO) following

Levin et al. (2010). Potassium salts are commonly observed in biomass smoke emissions

(e.g. P´ osfai et al., 2003; Freney et al., 2009; Semeniuk et al., 2007; Lewis et al.,

2009) and in fact potassium and chloride were the most abundant ions from the emissions

from most of these burns. Organic carbon was converted to POM by multiplying


OC by a molecular carbon to organic carbon multiplier.

The molecular carbon to organic carbon multiplier is necessary to account for other

elements associated with the organic carbon composition (Turpin and Lim, 2001). Values

can range from 1 to greater than 2 depending on sources and atmospheric processing.

(Turpin and Lim, 2001; Russell, 2003; El-Zanan et al., 2005; Malm and Hand,


2007). We derived estimates of the multiplier using a mass balance approach by forcing

closure between measured PM

2.5 gravimetric and reconstructed mass within 5 μgm3


<3% of mass). The values obtained ranged from 1.5 to 2.5, with an average

and one standard deviation of 1.6

±0.3, depending on the burn. While we recognize

these values are subject to uncertainties inherent in the mass balance approach (e.g.


El-Zanan et al., 2005; Malm and Hand, 2007), the derived estimates are consistent

with previous literature for biomass burning samples (Reid et al., 2005a; Malm et al.,

2005; Turpin and Lim, 2001). Further discussion of the multiplier and the sensitivity of


(RH) will be presented in Sect. 4.


2.3 Particle size distributions

Particle number concentrations as a function of size were measured using a di


mobility particle sizer (DMPS, TSI, Minneapolis, Minnesota) that included a di


mobility analyzer (TSI 3081) and associated condensation particle counter (TSI 3785).

A total of 24 bins were used with

5 a diameter range of 0.04<Dp <0.65 μm. The sample

was not dried because the RH of the burn chamber was typically lower than 20%.

More details regarding the size distribution measurements are provided by Levin et

al. (2010). From the 10 min dry aerosol number distributions we computed an average

size distribution over the entire (

&#24;2 h) burn. In some cases the size distributions evolved


over the burn duration, in other cases it remained fairly stable (see Levin et al., 2010).

Results from both McMeeking et al. (2009) and Levin et al. (2010) indicate that the

size distributions were dominated by accumulation mode particles. Sensitivity to the

variability of the size distribution to

f (RH) will be discussed in Sect. 4.

Mass concentrations were derived by integrating the volume size distributions mea


sured by the DMPS and multiplying by a mixture density derived from the chemical

composition data for each burn (see Sect. 3). Comparisons between DMPS reconstructed

mass and gravimetric fine mass from the IMPROVE samplers showed an

overestimation of the DMPS mass by a factor of six in some cases. Recall that the IMPROVE

sampler is a PM

2.5 sampler while the DMPS samples particles up to &#24;0.65 μm


in diameter, so it was expected that the DMPS would actually underestimate particulate

mass given the di

fferences in the upper size limits. Previous studies have shown

that when using the DMPS to sample non-spherical particles sizing discrepancies can

occur (e.g. Kr¨amer et al., 2000; Schneider et al., 2006; Khalizov et al., 2009). As

we will show in Sect. 4, many of the burns yielded fractal particles containing long


chains of soot as observed in scanning electron microscope (SEM) images. In fact,

the largest discrepancies in mass corresponded to samples dominated by aggregated

soot chains. To account for the e

ffects of non-sphericity on size distributions, we used

the dynamic shape factor, , defined as the ratio of the actual resistance force of a


non-spherical particle to the resistance force of a sphere having the same velocity and

volume (Hinds, 1999). For our application,

relates the equivalent diameter (Dp) to

the mobility diameter (

Dm) by Eq. (1):








A shape factor of =1 corresponds to a sphere and for irregularly shaped particles or

agglomerates it can be as high as 2 or more (Baron and Willeke, 1993). Shape factors

were derived by dividing the DMPS-derived mass by the IMPROVE gravimetric fine

mass and taking the cubed root. The values ranged from 0.8 to 1.8 with an average

and one standard deviation of 1.3

±0.3. These shape factors are in the range reported


by Kra¨mer et al. (2000) for sodium chloride crystals, and for biomass smoke (Schneider

et al., 2006; Gwaze et al., 2006). The highest values corresponded to the burns

with agglomerates of soot chains viewed by SEM images, consistent with observation

made earlier of the comparisons with the largest discrepancies in mass. There were

three cases when

<1 (ponderosa pine duff, Alaska duff and ceanothus); values less


than one are physically unrealistic and reflect the uncertainties in the estimates. For

example, values of

obviously were sensitive to the derived densities used to compute

mass, as well as any discrepancies in the volume or mass concentrations. We adjusted

the size distributions by dividing bin diameters by the shape factors derived for each

burn. Sensitivity of

f (RH) to the application of the shape factor will be discussed in


Sect. 4.

2.4 Scanning electron microscopy analysis

Particles were collected onto transmission electron microscope filmed grid substrates

(Carbon type B on Cu 400 mesh grids, Ted Pella Inc. Redding, California) for scanning

electron microscopy analysis using a rotating cascade impactor (MOUDI, model 110,


MSP, Inc.). Scanning electron microscopy imaging of collected samples and X-ray

microanalysis of particles were performed at the Environmental Molecular Sciences


Laboratory (Richland, WA), using a FEG XL30 digital scanning electron microscope

(FEI, Inc) equipped with an energy dispersed X-ray (EDX) spectrometer (EDAX, Inc).

Specific details of the SEM/EDX analysis of particles deposited onto filmed grid substrates

are described elsewhere (Laskin et al., 2006, and references therein).


2.5 Growth factors (GF)

Particle diameter growth factors (GF) were measured using a hygroscopic tandem


fferential mobility analyzer (HTDMA). The HTDMA includes a DMA that selects a

nearly-monodisperse aerosol sample (in this case, particles of 100nm mobility diameter),

followed by a conditioning system that subjects the particles to controlled water


sub-saturated environment (Rader and McMurry, 1986). A second DMA classifier measures

the size distributions of humidified particles. Measurements were made at eight

RH values ranging from 40% to 95%. A detailed description of the experimental protocol

is reported by Carrico et al. (2008, 2010). All experiments were conducted such

that particles were initially dry (RH

<15%) and then exposed to a preset higher RH be15

fore being measured by the second DMA. Growth factors are defined as the ratio of

the humidified diameter to the dry diameter. The estimated uncertainty in GF is 0.02

for spherical particles, combined with an estimated uncertainty in RH of 2% (Carrico et

al., 2010).

This entry was posted in C States = CA, CO, CT = California, Colorado, Connecticut. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s