Aerosol physical, chemical and optical properties during the Rocky Mountain
Airborne Nitrogen and Sulfur study
E.J.T. Levina,*, S.M. Kreidenweisa, G.R. McMeekinga, C.M. Carricoa, J.L. Collett, Jr.a, W.C. Malmb
aDepartment of Atmospheric Science, Colorado State University, 1371 Campus Delivery, Fort Collins, CO 80523-1371, USA
bNational Park Service/Cooperative Institute for Research in the Atmosphere, Colorado State University, Fort Collins, CO 80523-1375, USA
a r t i c l e i n f o
Received 4 September 2008
Received in revised form
18 December 2008
Accepted 18 December 2008
Remote aerosol concentrations
Remote aerosol composition
Rocky Mountain National Park air quality
a b s t r a c t
During the Rocky Mountain Airborne Nitrogen and Sulfur (RoMANS) study, conducted during the spring
and summer of 2006, a suite of instruments located near the eastern boundary of Rocky Mountain
National Park (RMNP) measured aerosol physical, chemical and optical properties. Three instruments,
a differential mobility particle sizer (DMPS), an optical particle counter (OPC), and an aerodynamic
particle sizer (APS), measured aerosol size distributions. Aerosols were sampled by an Interagency
Monitoring of Protected Visual Environments (IMPROVE) sampler and a URG denuder/filter-pack system
for compositional analysis. An Optec integrating nephelometer measured aerosol light scattering. The
spring time period had lower aerosol concentrations, with an average volume concentration of
2.2 ?2.6 mm3cm?3compared to 6.5 ?3.9 mm3cm?3in the summer. During the spring, soil was the single
largest constituent of PM2.5mass, accounting for 32%. During the summer, organic carbon accounted for
60% of the PM2.5mass. Sulfates and nitrates had higher fractional contributions in the spring than the
summer. Variability in aerosol number and volume concentrations and in composition was greater in the
spring than in the summer, reflecting differing meteorological conditions. Aerosol scattering coefficients
(bsp) measured by the nephelometer compared well with those calculated from Mie theory using size
distributions, composition data and modeled RH dependent water contents.
? 2008 Elsevier Ltd. All rights reserved.
The Rocky Mountain Airborne Nitrogen and Sulfur (RoMANS)
study was conducted at various sites throughout Colorado and
surrounding states in two phases during the spring and summer of
2006. This work focuses on aerosol data collected at the RoMANS
core site, located near the eastern boundary of Rocky Mountain
National Park (RMNP) 8 km south of Estes Park, Co. at an elevation
of 2750 m. The spring study period lasted from 23 March through
30 April and the summer period from 6 July though 12 August. The
overarching goals of RoMANS included characterizing wet and dry
deposition fluxes of sulfur and nitrogen in RMNP as well as iden-
tifying source types and regions for these species. Since dry depo-
sition of particulate matter and wet deposition of precipitation
scavenged particulate matter are contributors to N and S deposition
fluxes, characterization of aerosol concentrations and composition
is an important component of understanding pollutant deposition
fluxes in the region.
The particles that contribute to N and S deposition also affect
visibility. Since 1988 the InteragencyMonitoringof ProtectedVisual
Environments (IMPROVE) program has measured light scattering
aerosols in federally protected areas, such as RMNP (Malm et al.,
2004). In 1999 the Environmental Protection Agency enacted the
Regional Haze Rule which requires all federal Class I areas (national
parks and wilderness areas) to return visibility to natural condi-
tions within 60 years (Malm and Hand, 2007). In order to reach this
objective, it is imperative to know the species, and sources of these
species, that contribute to visibility degradation. The goal of this
paper is to examine aerosol concentration and speciation in RMNP
during the two RoMANS study periods and the effects these aero-
sols have on visibility in the park.
Three instruments located at the RoMANS core site measured
aerosol number size distributions with 15-min time resolution:
a differential mobility particle sizer (DMPS; TSI 3085), an optical
particle counter (OPC; PMS LASAIR 1003) and an aerodynamic
particle sizer (APS; TSI 3021). These three instruments measured
over the ranges 0.04–0.63 mm, 0.39–0.95 mm and 1.0–20 mm
* Corresponding author.
E-mail address: email@example.com (E.J.T. Levin).
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/atmosenv
1352-2310/$ – see front matter ? 2008 Elsevier Ltd. All rights reserved.
Atmospheric Environment 43 (2009) 1932–1939
respectively. Calibrations were checked twice during the spring
study period and three times during the summer period with
polystyrene latex (PSL) spheres of known sizes.
The sizing instruments were housed inside a temperature-
controlled mobile lab. Sample flow at 0.6 LPM was pulled through
an inlet in the roof and then passed through a Perma Pure dryer
(Perma Pure Inc., Toms River, NJ) which dried the sample to
RH<10% by supplying dry air to the sheath of the dryer. After the
dryer, the flow was split isokinetically intotwo0.3 LPM flows which
were sent tothe OPC and DMPS. Because the OPC operated at a flow
rate of only 0.028 LPM, the excess 0.272 LPM was siphoned off
before the OPC inlet using a critical orifice to control the flow. To
avoid losses of larger particles, the APS sampled through a separate
inlet, with no bends, at 5.0 LPM. The APS sample was dried using
a heating tape wrapped around the inlet tube which heated the
flow to w35?C, thus reducing RH<15%. The sample flow was
exposed to this heated region for about 0.5 s. This short residence
time limits the loss of volatile particles (An et al., 2007) yet is still
long enough for any water on the aerosol to react to the lower RH
(Snider and Petters, 2008). The inlets for both the DMPS/OPC and
APS were approximately 5 m above ground level.
A URG denuder and filter pack system collected 24-h samples to
determine concentrations of SO4
(2004), consists of a 2.5 mm aerodynamic size cut cyclone followed
in series by two annular denuders, a filter pack and then a third
denuder. The first denuder in the series was coated with Na2CO3
and the second with H3PO3to collect gas phase nitric acid and
ammonia, respectively. The nylon filter (Pall Corp. Nylasorb,1.0 mm
pore size) collected particulate matter and the third denuder
captured any volatilized NH4
retain any volatilized nitric acid (Yu et al., 2005). Samples were run
from 8:00 AM to 8:00 AM MST. Flow was controlled by a mass flow
controllerand verified byan in-line drygas meter. Aqueous extracts
of filters and denuders were analyzed by IC at Colorado State
University using methods similar to those outlined by Yu et al.
(2005). The URG sampling height was approximately 3 m above
The RoMANS main site was located at the RMNP IMPROVE
sampling site, which has been in place since 1988. The IMPROVE
sampler, described by DeBell et al. (2006), consists of four modules,
each with its ownpump, size cutting cyclone and sample substrate.
Three modules have a 2.5 mm aerodynamic size cut (PM2.5) and
collect aerosols onto Teflon, nylon and quartz filter substrates. The
fourth module uses a 10 mm aerodynamic size cut (PM10) and
collects particles onto a Teflon filter. The two Teflon filters are
weighed before and after sampling in a clean, climate controlled
room kept at an RH of 20–40%, to determine total gravimetric mass
loading. The PM2.5Teflon filter is also analyzed by X-ray fluores-
cence analysis for the common soil elements Al, Si, Ca, K, Fe, Ti, Mg
and Na. Concentrations of NO3
analysis of an aqueous extract of the nylon filter and the quartz
filter is analyzed via thermal optical reflectance (TOR) for organic
and elemental carbon. The analysis techniques are described in
greaterdetail by DeBell et al. (2006). During RoMANS, the IMPROVE
sampler was run on the same schedule as the URG sampler, daily
from 8:00 AM to 8:00 AM MST, which is different from the routine
network sampling schedule.
An Optec NGN-2 integrating nephelometer located at the
IMPROVE site measured ambient aerosol scattering. This instru-
ment has an open-air inlet, which does not deliberatelyexclude any
particles by size. The nephelometer uses a Lambertian light source
with an effective wavelength of 550 nm and detects scattered light
over the range 5?–175?(Optec NGN-2 Instrument Manual). An
internal sensor measures the RH of the sample in the inlet. The
?, Cl?, Naþ, Kþ, Ca2þ, Mg2þ
þions. The URG system, described in detail by Lee et al.
þ. The nylon filter has been shown to
2?are determined by IC
instrument was automatically calibrated with HEPA filtered air
every six hours and HFC-134A (SUVA) once a day.
2.1. Data analysis
Particle losses in the sizing instrument sampling lines and
Perma Pure dryer were calculated based on the procedure in
Hand (2001). Particle loss was generally below 10%, except for
particles in the largest OPC size bin, for which losses increased to
45%. Aerosol concentrations were adjusted to account for these
losses. Losses in the APS inlet were assumed to be negligible due
to its size and vertical geometry. Data from both the sizing system
and the Optec nephelometer were screened to remove spikes in
the data, defined as a greater than 100% difference between two
adjacent data points. The Optec data were also filtered to remove
times when RH>90%.
Because the three sizing instruments all measure over different
diameter ranges and exploit different aerosol characteristics, the
output from the instruments has to be reconciled to produce one
continuous size distribution. This was done following the align-
ment method developed by Hand and Kreidenweis (2002). This
technique was previously successfully employed during the Big
Bend Regional Aerosol and Visibility Observational (BRAVO) study
in 1999 (Hand et al., 2002) and the Yosemite Aerosol Character-
ization Study (YACS) in 2002 (McMeeking et al., 2005). Data from
the three instruments are first fit to a common diameter grid using
a Twomey fit (Winklmayr et al.,1990). The alignment method then
reconciles DMPS, OPC and APS data in two separate steps.
The first step fits the OPC data to the DMPS by adjusting the real
refractive index (n), used to define the particle size corresponding
to the OPCbins, since the OPCoutput is a function of this parameter.
OPC response to different n values was characterized by using size
selected aerosols of known n. By performing this test with oleic
acid, n¼1.46, ammonium sulfate, n¼1.53, and using the manu-
facture’s PSL calibrations, n¼1.588, OPC response curves can be
calculated for each bin (Hand and Kreidenweis, 2002). Table 1 gives
the lower bin limits for each OPC channel as a function of n.
The alignment process scans through n, from 1.400 to 1.600 in
increments of 0.005, and adjusts the OPC output based on the
previously calculated OPC response curves. The DMPS data are also
inverted in this step of the alignment. The DMA is run without an
impactor which would give a known upper size limit to sampled
particles. Instead, OPC data are assumed to represent singly
charged particles and are used to correct for multiply charged
particles in the DMA inversion (Hand and Kreidenweis, 2002). The
fit between the refractive index adjusted OPC and inverted DMPS
data is tested using the c2statistic and the refractive index leading
to the lowest c2is selected as the best-fit.
The APS measures aerodynamic diameter (Da) which is related
to equivalent spherical diameter (De) via the equation (Hinds,1999)
De ¼ Da
OPC lower bin limits as determined by calibration aerosols with different real
Channel PSL, m¼1.588 [mm] (NH4)2SO4, m¼1.53 [mm] Oleic acid, m ¼1.46 [mm]
2 0.2 0.241
3 0.3 0.363
4 0.4 0.548
5 0.5 0.730
6 0.7 0.810
E.J.T. Levin et al. / Atmospheric Environment 43 (2009) 1932–1939 1933
wherer0is 1.0 gcm?3. The second stepof the alignment fits the APS
data to the aligned OPC data by scanning through density from 1.20
to 3.00 g cm?3in 0.05 gcm?3increments. The APS output is con-
vertedtoequivalent spherical diameter using the assumed effective
density (re), which includes shape factor, and compared to the
aligned OPC data in the overlap region using c2, with the best-fit
density taken as that where c2is a minimum.
The final output from the alignment is a dry aerosol number
distribution between 0.04 and 20 mm expressed as dN/dlog10Dp
evaluated at 96 diameters with a base 10 logarithmic bin width of
0.03. The alignment program also records the real refractive index
and density resulting in the best-fit at each measurement point.
Hand et al. (2002), using the same technique, estimated uncer-
tainties of up to 10% in volume concentrations, 2% in accumulation
mode geometric mean diameter (Dgv) and 20% in coarse mode Dgv.
3. Aerosol size distributions
Volume distributions (dV/dlog10Dp) were calculated from the
number distributions and both number and volume distribution
statistics werecalculatedusing theequationsin Seinfeld and Pandis
(2006). We also calculated integrated total number and volume
concentrations. Table 2 lists the mean and one standard deviation
for total aerosol number and volume concentrations as well as
accumulation and coarse mode concentrations and mode statistics.
While measured aerosol number concentrations were generally
low in RMNP, they were significantly higher during the summer
than the spring. The mean, and one standard deviation, aerosol
number concentration in the spring was 880?770 cm?3while in
the summer it was 2080 ?940 cm?3and total integrated volume
concentrations were 2.2?2.6 mm3cm?3
6.5?3.9 mm3cm?3in the summer. Although these standard devi-
ations are large, aerosol concentrations (both number and volume)
were significantly different between spring and summer at the 95%
confidence level using a two tailed t-test.
Aerosol distributions were split into accumulation and coarse
modes based on the minimum values between the modes in the
volume distributions. This minimum occurred at an average value
of 0.63 mm in the spring and 0.74 mm in the summer. In the spring,
the coarse mode dominated the volume distribution, accounting
for 60% of the total aerosol volume. In the summer, however, the
accumulation mode made up 60% of the aerosol volume.
Geometric mean diameters (Dg) and standard deviations (sg)
were calculated for each number (subscript n) and volume
(subscript v) distribution. The accumulation mode Dgvremained
constant throughout both study periods with an average value of
0.2?0.03 mm. For the coarse mode, however, Dgv increased
4.7?0.9 mm comparedto 3.4?1.3 mm in the spring. We note that, if
the particle number distributions were cut off at 10 mm, as would
occur in a PM10sampler, the computed mean Dgvfor the coarse
mode decreased: spring, 2.6 mm, and summer, 3.2 mm, indicating
that the PM10sampler missed some fraction of the coarse mode
aerosol. In the spring, the average sgvfor the accumulation and
whenits average valuewas
coarse modes were 1.8?0.1 mm and 2.0? 0.2 mm, respectively. In
the summer, both modes of the volume distribution were narrower
with sgvof 1.6?0.07 for the accumulation mode and 1.9?0.1 for
the coarse mode.
In addition to having lower aerosol concentrations, the spring
also exhibited greater variability. The number concentration varied
88% about its mean during the spring and only 42% during the
summer. Percent variations about the mean for the volume
concentration, coarse mode Dgv, and sgvfor both modes were also
larger in the spring than in the summer.
4. Aerosol composition
Data from the IMPROVE and URG samplers were used to
determine PM2.5 aerosol composition. Ammonium, sulfate and
nitrate concentrations were taken from the URG data while
IMPROVE data were used for all other species. Malm et al.
(submitted for publication) discuss the comparisons between
duplicate measurements from the two samplers. While SO4
concentrations were in generally good agreement, NO3
trations differed by 15–20%, and the IMPROVE NH4
were roughly 50% less than those from the URG sampler. The
discrepancies are likely due to the lack of front-end and backup
denuders on the IMPROVE sampler, and thus the URG data are
considered to be more reliable. A charge balance was performed
with the URG data (Fig. 1). The ratio of cations to anions was
0.9?0.2 in the spring and 1.2?0.2 in the summer. The slightly high
value in the summer indicates an excess of measured ammonium,
which could be accounted for by oxalate. We did not analyze
samples for oxalate; however, its mass is expected to be accounted
for in the bulk organic carbon measurement, so we do not sepa-
rately compute it. Since it is a minor component, it has no
discernable effect on any of our following calculations.
Following standard IMPROVE procedures (Pitchford et al., 2007),
we assumed that reconstructed fine mass (RCFM) was completely
accounted for by the following components:
½RCFM? [ ½OMC?D½SOIL?D½SULFATE?D½NITRATE?D½LAC?
where the terms are the mass concentrations of organics, soil,
sulfates, nitrates and light absorbing carbon respectively. We have
neglected the contribution of sea salt, which was insignificant. Each
of these components is computed as follows.
Total organic mass concentrations [OMC] were computed by
multiplying the mass concentration of organic carbon [OC] by
a coefficient to account for the mass of other elements, such as H, O
and N, in the organic molecule. The standard IMPROVE algorithm
uses an [OC] coefficient of 1.8 for every site (Pitchford et al., 2007).
There can be great variability in organic molecules, however, and
Malm and Hand (2007) found that an [OC] multiplier of 1.7 worked
best for the RMNP IMPROVE site. We also found that a value of 1.7
gave the best agreement for the spring data, and thus applied the
Mean and standard deviations for number and volume aerosol concentrations and mode statistics. Concentration units are: number [cm?3], volume [mm3cm?3].
E.J.T. Levin et al. / Atmospheric Environment 43 (2009) 1932–19391934
In the summer, using a coefficient of 1.7 underestimated the
gravimetric mass. To achieve closure between gravimetric and
reconstructed fine mass in summer, [OC] was multiplied by 1.95.
Soil mass concentrations were computed using:
½SOIL? [ 2:2½Al?D2:49½Si?D1:63½Ca?D2:42½Fe?D1:94½Ti?
where the terms are mass concentrations of aluminum, silicon,
calcium, iron and titanium and the coefficients account for the
common oxides of these soil elements (Malm et al., 2004). These
coefficients are all consistent with standard IMPROVE assumptions.
In contrast to the standard IMPROVE assumption, we did not
compute total sulfate compound mass assuming the sulfate ionwas
fully ammoniated. Sulfate ammoniationwas allowed tovaryamong
measured ratio of moles of ammonium to moles of sulfate (Fig. 2).
In the spring, the mole ratio varied between 1 and 5 so it was
assumed that the various forms of ammoniated sulfate were
present in the appropriate proportions for each day. In summer, the
molar ratio was always greater than or equal to two, so only
(NH4)2SO4was assumed to be present.
Light absorbing carbon mass concentration [LAC] was equated
to that of elemental carbon determined from the TOR analysis.
To test the assumption that RCFM completelyaccounts for PM2.5
mass, as well as the assumptions in calculating the RCFM compo-
nents, RCFM was plotted against the measured PM2.5gravimetric
mass (GM). These data showed very good agreement using the
above assumptions with an R2value of 0.96 in the spring and 0.98
in the summer and a regression line slope of 1.0 for both seasons.
Averaged over each study period, soil was the single largest
constituent of PM2.5in the spring, while organic carbon dominated
PM2.5in the summer. Also, both sulfate and nitrate contributed
significantly more to aerosol mass in the spring than the summer.
Similar to what was seen above with aerosol size distributions,
spring composition also had greater variability with large spikes in
soil concentration on the 9th,16th and 22nd of April (Fig. 3). These
spikes in soil may be the result of long range Asian dust transport
which often affects thewestern US in the spring (Malm et al., 2004),
although the transport patterns associated with Asian dust also
depending on the
frequently move through the southwestern U.S. and can pick up
dusts from these regions (Richardson et al., 2007). In the summer,
aerosol composition remained more constant, with organics
accounting for over half of PM2.5throughout most of the study
period. Biomass burning can be a significant source of organic
species and often affects aerosol composition and loading
throughout the western US during the summer months (Park et al.,
Table 3 shows the average, and one standard deviation, mass
fractions of each RCFM component during the two time periods.
The same table also shows the average mass fractions from the
IMPROVE data for the months of April and July averaged over
1991–2006. In calculating the historical aerosol composition, we
used the same [OC] coefficients as used for the RoMANS data, but
assume ammonium sulfate. Aerosol composition during the
RoMANS spring period was very similar to the historical data with
no component showing a significant difference from the historical
average. As with the RoMANS summer data, OMC dominated
aerosol composition in the historical data, although its mass frac-
tion was slightly lower than that measured during RoMANS. Also,
during the summer, the fraction attributed to ammoniated sulfate
was lower than that in the historical record. Average, and standard
deviation, daily PM2.5 mass loading during the two RoMANS
periods were 3.1?1.4 and 5.6?2.2 mgm?3. Historical values for
these two time periods are 3.4?1.7 and 5.3?2.5 mgm?3.
We calculated dry aerosol refractive index (mcomp) and density
(rcomp) from the daily composition data, assuming internal mixing,
using (Hasan and Dzubay, 1983):
þis not measured in the network, the historical data
mcomp ¼ rcomp
where Xi,ri, niand kiare respectively the mass fraction, density, real
refractive index and imaginary refractive index for species i. Table 4
gives values for the individual species densities and refractive
indices used here.
Mean and one standard deviation PM2.5mass fractions for aerosol components
during the two RoMANS periods as well as historical means and standard deviations
for April and July from 1991 to 2006.
AS ANOMC LACSOIL
Historical 0.31 ?0.06 0.15?0.05 0.26?0.06 0.03?0.01 0.24?0.09
Summer RoMANS0.16 ?0.04 0.03?0.02 0.60?0.12 0.04?0.01
Historical 0.24?0.05 0.05?0.02 0.51?0.06 0.04?0.01
0.24?0.08 0.13?0.10.29?0.09 0.03?0.02 0.32 ?0.14
Mar 25 Apr 08Apr 22
Jul 15Jul 29Aug 12
Fig. 3. Mass fraction of PM2.5components: ammoniated sulfate (red), ammonium
nitrate (blue) organic carbon (green), elemental carbon (black) and soil (brown). (For
interpretation of the references to color in this figure legend, the reader is referred to
the web version of this article.)
Mar 25Apr 08Apr 22
cations / anions
Jul 15 Jul 29Aug 12
Fig. 1. Ratio of cations to anions from the URG data in equivalent units. Grey line
indicates charge balance.
Apr 08Apr 22
MolesNH4 / MolesSO4
Jul 15Jul 29Aug 12
Fig. 2. Ratio of moles of ammonium to moles of sulfate. Times with points above the
grey line are assumed to have fully ammoniated sulfate.
E.J.T. Levin et al. / Atmospheric Environment 43 (2009) 1932–19391935
Fig. 4 shows the real (ncomp) and imaginary (kcomp) dry refractive
indices calculated from composition for both study periods plotted
in red and blue respectively. The black line is the best-fit refractive
index retrieved for the dry aerosol by the alignment method
(mretrieved). The average refractive index calculated from composi-
tion data was ncomp¼1.59?0.016, kcomp¼0.020? 0.012 for the
spring and ncomp¼1.57?0.009, kcomp¼0.023?0.004 for the
summer. The average retrieved refractive indices were in good
agreement with these estimates, with mretrieved¼1.58?0.011 and
mretrieved¼1.56?0.011 for the two study periods.
5. Data intercomparison
5.1. Mass/volume comparison
Aerosol density was estimated from IMPROVE PM2.5gravimetric
mass concentrations and aligned PM2.5volume concentrations. The
slope of the zero-intercept regression line through these points
gives an estimate of the aerosol density for each time period:
r¼1.96?0.19 g cm?3for the spring and r¼1.40? 0.06 gcm?3for
the summer. The spring density agrees with that calculated from
composition, rcomp¼1.9?0.15 gcm?3, within the uncertainty.
However, the summer aerosol density estimated by the above
method is significantly less than that calculated from composition,
rcomp¼1.62?0.11 gcm?3. Agreement within the uncertainty can
be achieved, however, if the density of organics (roc) in summer-
time is assumed to be 1.2 gcm?3instead of 1.4 gcm?3as listed in
Table 4. This results in an average composition derived density of
rcomp¼1.46?0.11 gcm?3. We note that measurement and model
assumption uncertainties and biases also contribute to the
observed slope between volume and mass concentrations, and the
adjustment of the average densities is only one way to explain this
slope. However, the requirement for a different OC-to-OMC
conversion factor to achieve mass closure in summertime suggests
it is not unreasonable to expect other organic aerosol properties to
also differ between these seasons. Both of these assumed rocvalues
are within the ranges cited by Turpin and Lim (2001), and induce
only minor changes (2% or less) in the computed optical properties.
The R2value for the mass/volume comparison is much larger in the
summer, 0.96 than the spring, 0.66, indicating aerosol density
varied more in the spring than in the summer, consistent with the
more variable composition in the spring.
5.2. Scattering comparison
Aerosol scattering coefficients can be calculated using Mie
theory if aerosol size distribution and refractive index are known
and if spherical particles are assumed. We used mcompvalues to
calculate dry light scattering efficiencies (Qscat) for each alignment
bin midpoint diameter. We assumed that the aerosol in the fine
mode was internally mixed, and that the fine mode aerosol
composition was that of the measured PM2.5fraction. The coarse
mode was assumed to contain only soil. Because composition data
have only daily temporal resolution, we assumed that these values
apply from 8:00 AM to 8:00 AM each day. Using the calculated Qscat
values, we then calculated dry aerosol scattering coefficients (bsp)
in Mm?1using the following equation (Seinfeld and Pandis, 2006)
where Dp,iis the bin midpoint diameter [mm], Niis particle number
concentration [cm?3], and the sum is over all bins. The calculation
was done for a wavelength of 550 nm, and for each 15-min size
distribution sampling period, initially neglecting any changes in
the measured dry aerosol size distributions due to water uptake.
The calculated bspvalues, as well as ambient bspmeasured by the
nephelometer, were averaged to an hourly time interval for
comparisons. The nephelometer data were not adjusted to account
for any measurement nonidealities.
In the spring, the calculated dry bspunderestimated measured
bsp, especially for the higher values. There was much better
agreement during the summer. The slope of the regression line
through the calculated versus measured data points, using simple
linear regression, is 0.84 in the summer compared with 0.4 in the
spring (Table 5). The summer data are also more highly correlated,
with an R2¼0.82 while for the spring R2¼0.59.
Much of the discrepancy between the measured and calculated
bspvalues in the spring can be accounted for by the effects of
humidity. Aerosol water uptake affects both particle size and
refractive index and can have a large effect on bsp(Malm and Day,
Densities and refractive indices for species used in reconstructed fine mass
Species Density [gcm?3]Refractive
Mar 25Apr 08 Apr 22
Jul 15Jul 29Aug 12
Fig. 4. Daily averaged effective refractive index retrieved from alignment method
(black) and real (red) and imaginary (blue) refractive indices calculated from aerosol
composition. (For interpretation of the references to color in this figure legend, the
reader is referred to the web version of this article.)
Average and standard deviation values for bsp measured by the Optec and bsp
calculated using dry aerosols (Miedry), wet accumulation mode aerosols (Miewet_acc)
and wet accumulation plus coarse mode aerosols (Miewet_total) [Mm?1]. Regression
statistics, from simple linear regression, show the agreement between calculated
and measured values for each case.
E.J.T. Levin et al. / Atmospheric Environment 43 (2009) 1932–19391936
2001). To account for water uptake we used the online Aerosol
Inorganics Model (AIM) (Wexlerand Clegg, 2002) to generatewater
uptake curves as a function of relative humidity for each of the
assumed nitrate and sulfate species. The AIM code was run in
metastable mode at 298.15 K. Relative humidity was scanned from
0.1 to 0.99 and polynomial fits constructed. We used these curves to
calculate the volume of water associated with each species at every
time point for the ambient RH measured by the Optec nephelom-
eter, assuming that water activity is equivalent to RH, and then
calculated total water associated with the aerosol using the ZSR
assumption (Stokes and Robinson,1966). We assumed that soil and
carbonaceous species had zero water uptake (Carrico et al., 2005).
This new composition was then used to calculate the aerosol sizes
and refractive indices used by the Mie code to reconstruct
As can be seen in Table 5, including aerosol water uptake in the
calculated bsp(Miewet_total) improves the agreement with measured
bspfor both seasons. In the spring, the slope of the regression line
increased to 0.66 and in the summer to 0.96. Also, the R2values for
both seasons increased, 0.87 in the spring and 0.90 in the summer,
when water uptake was accounted for. We note that our calcula-
tions show that our choices for model inputs produced estimates of
scattering coefficients that were consistent with the measure-
ments, but that alternative assumptions – e.g., regarding the mixing
of constituents within individual particles, the water uptake
properties of organic carbon, and other assumptions – could also be
Based on our calculations, aerosol water uptake played a much
greater role in bspduring the spring than the summer. During the
spring, calculated bspincreased by 20% on average when water was
included. For the highest bspvalues, calculated scattering almost
doubled. These high scattering points (indicated in grey in Fig. 5a)
are all from April 24 to 25 when RH was high, greater than 80%, and
nitrate and sulfate species accounted for roughly 40% and 25%,
respectively, of PM2.5 mass concentrations (See Fig. 3). These
species have large water concentrations associated with them for
RH>80% and therefore lead to much higher bspvalues. However,
evenwith the drastic increase in calculated bspthese values are still
lower than the measured bspvalues, leading to the regression line
slope less than 1. Because these points all fall during the time
period with the highest RH, this could indicate a slight bias in the
RH measurements, as small errors lead to larger discrepancies at
higher RH values. The underprediction of bsp during this time
period could also indicate that our assumed composition and/or
assumed mixing state are incorrect during this time. We note that
the data show no evidence of the presence of highly hygroscopic
sea salt particles, and the coarse mode volume fraction was low,
about 10%, so errors in the assumptions regarding the contributions
to bspfrom those factors are probably not important in explaining
the discrepancies. If these points are removed, calculated and
measured bspshow very good agreement when water uptake is
included (red line in Fig. 5a).
In the summer, the differences between wet and dry scattering
coefficients are much smaller, with an average increase of only 6%.
The differences between the two seasons are due to aerosol
composition and not differences in humidity. The average humidity
was actually higher in the summer, 51% versus 41%, and the
humidity was greater than 60% more often in the summer, 27% of
the time, than in the spring, 17% of the time. This difference in
aerosol hygroscopicity between the spring and summer is consis-
tent with seasonal aerosol composition. In the summer the mean
24-h composition was dominated by carbonaceous material with
little or no hygroscopic growth.
In Table 5 Miewet_accrefers to bspcalculated using only humidi-
fied accumulation mode aerosols. By comparing these values to
Miewet_totalwe can estimate the contribution of the coarse mode to
total scattering. In both seasons scattering was dominated by
accumulation mode aerosols; however, coarse mode aerosols
contributed more to scattering in the spring than the summer.
Averaged over each season, coarse mode aerosols accounted for 9%
of the total scattering in the spring and 5% in the summer. Again,
this is consistent with previous results which showed that
the accumulation mode dominated the aerosol volume during the
summer, but coarse mode accounted for a greater fraction of
the volume in the spring.
Optec Measured bsp [Mm-1]
Size Distribution Calculated bsp [Mm-1]
R2 = 0.87
y = 0.66x +2.06
R2 = 0.90
y = 0.96x +2.26
R2 = 0.89
y = 0.89x +0.50
Optec Measured bsp [Mm-1]
Size Distribution Calculated bsp [Mm-1]
Fig. 5. Calculated versus measured bsp[Mm?1] using humidified accumulation and
coarse mode aerosols during (a) spring (b) summer. Grey points in panel A indicate the
time period April 24–25. Solid lines are regression lines, dotted line is 1:1 line. The red
regression line in panel A does not include the grey points (see text). (For interpre-
tation of the references to color in this figure legend, the reader is referred to the web
version of this article.)
E.J.T. Levin et al. / Atmospheric Environment 43 (2009) 1932–19391937
measured at a remote monitoring station in Rocky Mountain
National Park during the spring and summer of 2006. Measure-
ments included aerosol light scattering, PM2.5chemical composi-
tion, and aerosol particle sizing using electrical mobility, optical
sizing, and time of flight techniques.
Aerosol concentrations were significantly lower during the
spring study period than during the summer. The spring also
exhibited much greater variation in aerosol concentration and size
distribution mode statistics. Overall, aerosol concentrations were
lower than those measured in other western National Parks (Hand
et al., 2002; McMeeking et al., 2005). Average summer accumula-
6.7?3.9 mmcm?3were measured using similar techniques at Big
Bend and Yosemite National Parks respectively. The accumulation
mode Dgvwas also slightly smaller in RMNP compared to Big Bend,
0.26?0.04 mm, and Yosemite, 0.28?0.05 mm.
As with aerosol concentration, aerosol composition exhibited
greater variability during the spring. Averaged over the entire
spring study period, soil was the largest constituent of PM2.5mass.
However, sulfates, nitrates and organics also contributed signifi-
cantly to aerosol composition during the spring and at times were
the dominant components. Throughout the entire summer study
period, however, organics dominated PM2.5composition.
By plotting PM2.5gravimetric mass concentration versus PM2.5
volume concentration we estimate that the seasonal average
aerosol densities were r¼1.96?0.19 gcm?3during the spring and
r¼1.40? 0.06 gcm?3during the summer. The average spring
value agrees with that calculated from composition data within the
uncertainties. However, the low correlation between mass and
volume concentrations during the spring, R2¼0.66, indicates that
assuming a single density for the entire period may be a poor
assumption. In the summer, agreement between estimated density
and that calculatedfromcomposition can be achieved if a densityof
1.2 is assumed for organics. The data are more highly correlated in
the summer, R2¼0.96, reflecting the lower variability in aerosol
composition during this time.
The season averaged refractive indices retrieved by the align-
1.56?0.011 for spring and summer respectively. These agree with
the real refractive indices calculated from composition data within
During the spring, the inclusion of aerosol water uptake
changed the calculated bspvalues by 20% on average and in some
cases 50%. Including the effects of water uptake improves closure
between measured and calculated bspin the spring, especially for
low bspvalues. During the summer, the effects of including water
uptake in the scattering calculations are much smaller, 6% on
average, despite similar ambient RH values, indicating that the
hygroscopicity of the summertime aerosol plays less of a role in
total ambient light scattering.
Overall, the low aerosol concentrations and bspmeasured in the
spring indicate that visibility degradation is not a major concern in
RMNP during this time. However, on the days with the highest bsp
the RH was high and the PM2.5aerosol composition was dominated
by hygroscopic species. Because of the variable nature of aerosol
composition and concentration, any visibility degradation in the
park during the spring is episodic and likely related to larger scale
In the summer, higher aerosol concentrations led to higher bsp
and thus a decreased visible range. The park also receives the
majority of its visitors during this time, so visibility in the summer
is a greater concern. Unlike in the spring, organics, not hygroscopic
physical,chemical andoptical propertieswere
salts, were the primary aerosol species during the summer, and the
lower hygroscopicity of the organic species led to less sensitivity of
visibility degradation to ambient RH. Instead, visibility was more
affected by total aerosol loading. Measured bspis highest in the
summer when total aerosol loading is highest.
The assumptions, findings, conclusions, judgments, and views
presented herein are those of the authors and should not be
interpreted as necessarily representing the National Park Service
We gratefully acknowledge logistical support for the RoMANS
study by Air Resource Specialists, Inc as well as Judy Vistyand Laura
Wheatley at RMNP. The authors also thank Taehyoung Lee, Amy
Sullivan, Suresh Raja and Florian Schwandner for the URG data.
Funding for this work was provided by the National Park Service
through Cooperative Agreement # H2380040002 and from the
State of Colorado.
An, W.J., Pathak, R.K., Lee, B.H., Pandis, S.N., 2007. Aerosol volatility measurement
using an improved thermodenuder: application to secondary organic aerosol.
Journal of Aerosol Science 38 (3), 305–314.
Carrico, C.M., Kreidenweis, S.M., Malm, W.C., Day, D.E., Lee, T., Carrillo, J.,
McMeeking, G.R., Collett, J.L., 2005. Hygroscopic growth behavior of a carbon-
dominated aerosol in Yosemite National Park. Atmospheric Environment 39 (8),
DeBell, L.J., Gebhart, K.A., Hand, J.L., Malm, W.C., Pitchford, M.L., Schichtel, B.A.,
White, W.H., 2006. Spatial and Seasonal Patterns and Temporal Variability of
Haze and its Constituents in the United States. Report IV. Cooperative Institute
for Research in the Atmosphere, Fort Collins, CO. http://vista.cira.colostate.edu/
Dick, W.D., Saxena, P., McMurry, P.H., 2000. Estimation of water uptake by organic
compounds in submicron aerosols measured during the Southeastern Aerosol
and Visibility Study. Journal of Geophysical Research – Atmospheres 105 (D1),
Hand, J.L., 2001. A New Technique for Obtaining Aerosol Size Distributions with
Applications to Estimates of Aerosol Properties. Ph.D. dissertation, Colorado
State University, Fort Collins, CO.
Hand, J.L., Kreidenweis, S.M., 2002. A new method for retrieving particle refractive
index and effective density from aerosol size distribution data. Aerosol Science
and Technology 36 (10), 1012–1026.
Hand, J.L., Kreidenweis, S.M., Sherman, D.E., Collett, J.L., Hering, S.V., Day, D.E.,
Malm, W.C., 2002. Aerosol size distributions and visibility estimates during the
Big Bend regional aerosol and visibility observational (BRAVO) study. Atmo-
spheric Environment 36 (32), 5043–5055.
Hasan, H., Dzubay, T.G.,1983. Apportioning light extinction coefficients to chemical-
species in atmospheric aerosol. Atmospheric Environment 17 (8), 1573–1581.
Hinds, W.C., 1999. Aerosol Technology: Properties, Behavior, and Measurement of
Airborne Particles. Wiley, New York.
Lee, T., Kreidenweis, S.M., Collett, J.L., 2004. Aerosol ion characteristics during the
Big Bend Regional Aerosol and Visibility Observational study. Journal of the Air
& Waste Management Association 54 (5), 585–592.
Lide, D.R., 2008. CRC Handbook of Chemistry and Physics [A Ready-reference Book
of Chemical and Physical Data]. CRC Taylor & Francis.
Malm, W.C., Day, D.E., 2001. Estimates of aerosol species scattering characteristics as
a function of relative humidity. Atmospheric Environment 35 (16), 2845–2860.
Malm, W.C., Hand, J.L., 2007. An examination of the physical and optical properties
of aerosols collected in the IMPROVE program. Atmospheric Environment 41
Malm, W.C., McMeeking, G.R., Kreidenweis, S.M., Levin, E.J.T., Carrico, C.M., Day, D.E.,
Collett, J.L., Lee, T., Sullivan, A.P., Raja, S. Using high time resolution aerosol and
number size distribution measurements to estimate atmospheric extinction.
Journal of the Air & Waste Management Association, submitted for publication.
Malm, W.C., Schichtel, B.A., Pitchford, M.L., Ashbaugh, L.L., Eldred, R.A., 2004. Spatial
and monthly trends in speciated fine particle concentration in the United
States. Journal of Geophysical Research – Atmospheres 109 (D3).
McMeeking, G.R., Kreidenweis, S.M., Carrico, C.M., Lee, T., Collett, J.L., Malm, W.C.,
2005. Observations of smoke-influenced aerosol during the Yosemite Aerosol
E.J.T. Levin et al. / Atmospheric Environment 43 (2009) 1932–19391938
Characterization Study: size distributions and chemical composition. Journal of
Geophysical Research – Atmospheres 110 (D9).
Park, R.J., Jacob, D.J., Logan, J.A., 2007. Fire and biofuel contributions to annual mean
aerosol mass concentrations in the United States. Atmospheric Environment 41
Pitchford, M., Malm, W., Schichtel, B., Kumar, N., Lowenthal, D., Hand, J., 2007.
Revised algorithm for estimating light extinction from IMPROVE particle
speciation data. Journal of the Air & Waste Management Association 57 (11),
Richardson, M.S., DeMott, P.J., Kreidenweis, S.M., Cziczo, D.J., Dunlea, E.J.,
Jimenez, J.L., Thomson, D.S., Ashbaugh, L.L., Borys, R.D., Westphal, D.L.,
Casuccio, G.S., Lersch, T.L., 2007. Measurements of heterogeneous ice nuclei in
the western United States in springtime and their relation to aerosol charac-
teristics. Journal of Geophysical Research – Atmospheres 112 (D2).
Seinfeld, J.H., Pandis, S.N., 2006. Atmospheric Chemistry and Physics: From Air
Pollution to Climate Change. Wiley, Hoboken, NJ.
Snider, J.R., Petters, M.D., 2008. Optical particle counter measurement of marine
aerosol hygroscopic growth. Atmospheric Chemistry and Physics 8 (7),
Stelson, A.W., 1990. Urban aerosol refractive-index prediction by partial molar
refraction approach. Environmental Science and Technology 24 (11),1676–1679.
Stokes, R.H., Robinson, R.A., 1966. Interactions in aqueous nonelectrolyte solutions.
I. solute–solvent equilibria. Journal of Physical Chemistry 70 (7), 2126–2131.
Tang, I.N.,1996. Chemical and size effects of hygroscopic aerosols on light scattering
coefficients. Journal of Geophysical Research Atmospheres 101 (D14), 19245–
Turpin, B.J., Lim, H.J., 2001. Species contributions to PM2.5mass concentrations:
revisiting common assumptions for estimating organic mass. Aerosol Science
and Technology 35 (1), 602–610.
Wexler, A.S., Clegg, S.L., 2002. Atmospheric aerosol models for systems including
the ions Hþ, NH4
Research – Atmospheres 107 (D14).
Winklmayr, W., Wang, H.C., John, W., 1990. Adaptation of the Twomey algorithm to
the inversion of cascade impactor data. Aerosol Science and Technology 13 (3),
Yu, X.Y., Taehyoung, L., Ayres, B., Kreidenweis, S.M., Collett, J.L., Maim, W., 2005.
Particulate nitrate measurement using nylon filters. Journal of the Air & Waste
Management Association 55 (8), 1100–1110.
þ, Naþ, SO4
?, Cl?, Br?, and H2O. Journal of Geophysical
E.J.T. Levin et al. / Atmospheric Environment 43 (2009) 1932–1939 1939