Size-dependent aerosol deposition velocities during BEARPEX`07

1
in prep for Atmospheric Chemistry and Physics
Size-dependent aerosol deposition velocities during BEARPEX’07
6-2-09
Richard J. Vong1, Ivar J. Vong1, Dean Vickers1, David S. Covert2
1
College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis OR 97331 USA
Joint Institute for the Study of Atmospheres and Oceans, University of Washington, Seattle WA
,USA
2
Abstract
Aerosol concentrations and 3-D winds were measured from September 9 to 25, 2007, above a pine
forest in California. The measurements were combined using the eddy correlation (EC) technique to
determine aerosol eddy fluxes as a function of particle diameter within the accumulation mode size
range (0.25 µm < dia. < 1 µm here). The measured heat and water vapor fluxes were utilized to
correct the aerosol eddy fluxes for aerosol hygroscopic growth and vertical variations in density
were evaluated and removed from the data. The hygroscopic growth correction was necessary despite
the low RH and fairly hygrophobic nature of the particles. Uncertainties associated with particle
counting were evaluated from the data. Aerosol deposition velocities (Vd ) were shown vary from 0.2
to 1.0 cm/sec; Vd increases with friction velocity and particle diameter.
1 Introduction
Removal of aerosol to vegetation by atmospheric turbulence is the focus of an experiment that
addressed key factors which control the magnitude of aerosol flux and its uncertainties. These factors
relate to aerosol microphysics, aerosol chemical composition, and boundary layer dynamics.
The concentration and chemical composition of atmospheric accumulation mode (0.1 µm < diameter
< 2.0 µm) aerosol are important influences on: the Earth’s energy budget and global climate
(Charlson et al., 1991; 1992); air quality and pollution; and clouds and precipitation (Charlson et
al.,1987). The composition, concentration, and spatial distribution of aerosol are controlled by
emission, transport and mixing, conversion, and deposition processes. Aerosol removal in
precipitation is known as ‘wet deposition’ while removal by turbulence, impaction, diffusion and
gravitational settling are termed ‘dry deposition’.
Dry deposition can contribute a substantial fraction (up to one-half) of the total chemical deposition
(Erisman et al.,1997; Hicks et al. ,1991; Langer and Rodhe,1991; Balkanski et al., 1993; Marner and
Harrison, 2004) and can result in potentially significant impacts on terrestrial and aquatic
ecosystems.
Pryor et al. 2008; Gallagher, Nemitz, Fairall, Businger
2 Methods
2.1 Site
Aerosol concentrations and winds were measured from September 9 to 25, 2007, above a pine forest
owned by Sierra Pacific Industries, adjacent to the University of California at Berkeley’s Blodgett
Forest Research Station as part of the Biosphere Effects on Aerosols and Photochemistry
Experiment (BEARPEX). The site is located near Georgetown, CA (38 o 59’ N, 120 o 58’W) at 1315
m. elevation on the western slope of the Sierra Nevada Mountains and 75 km NE of Sacramento.
The area was planted with Pinus ponderosa in 1990 with a few other species present; average canopy
2
height was 7.9 m. The understory is composed of Manzanita and whitehorn shrubs up to 2 m in
height (displacement z= and. . The leaf area index for the ful canopy was 5.1 m 2 /m 2 . A detailed
description of the site is provided by Goldstein (2000). The daytime fetch is excellent in that the
upwind canopy is even aged and uniform, and the terrain is gently sloped (2 o ) for 2 km upwind
(winds from the SE to W). The night time fetch (winds from the E and N) is not as good, as
discussed further below.
1.2 Instrumentation
Aerosol concentrations as a function of particle diameter were measured by light scattering
techniques. The “FAST” aerosol spectrometer (Droplet Measurement Technologies, Boulder CO)
provided particle concentration at 10 Hz as a function of size for the EC fluxes while two identical
optical particle counters (Weslas OPC) measured the aerosol size spectra at 5 min. resolution; the
FAST and OPCs provided detailed resolution of the aerosol size spectrum for much of the
accumulation mode size range (0.25 µm < dia. < 1.0 µm).
The FAST laser passes through the aspirated aerosol before detection of the scattered light (5 o to
14 o forward collection angle; λ=680 nm) by masked and signal detectors (Vong et al. 2003). The
FAST was size-calibrated using PSL beads and its counting efficiency as a function of diameter was
quantified by comparison to a second instrument (DMT UHSAS) in the lab and to the OPCs in the
field. The FAST undercounted the smaller particles (dia < 0.4 µm) compared to the UHSAS and
OPCs before correction. The FAST operated continuously during daytime and some nights.
The two OPCs were operated for a few hours each day to quantify aerosol hygroscopic growth during
both morning and afternoon periods. Sample air was brought to relative humidities (RH) that spanned
the ambient RH in order to evaluate the hygroscopic growth factor (γ). The OPCs were located
inside a sampling van at the base of the tower and drew air through a common ½ cm dia. stainless
steel tube that originated at the 18.8 m AGL tower level near the EC instrumentation.
Water vapor was measured in situ at 10 Hz by ultraviolet absorption (Camp. Sci. KH-20), three
dimensional winds and virtual temperature at 10 Hz using an ultrasonic anemometer (Appl. Tech.
Inc., SWS-211-3K). as well as temperature and RH (Campbell HMP-45C) gradients at 30 minute
intervals (at 18.8 m / 7.3 m AGL).
The EC technique combined aerosol concentrations from the FAST, water vapor density from the
KH-20, and 3-D winds from the sonic anemometer as calculated covariances; all of these
measurements were performed from the top of a scaffolding tower at 18.8 meter AGL. The EC
instruments were mounted on a boom that extended 2 m upwind of the SW tower corner in order to
minimize flow distortion from the scaffolding. The boom and instruments were periodically oriented
(generally 2 to 3 times per day) into the prevailing wind direction by rotating the boom.
3. Data processing for aerosol EC flux es
The eddy correlation (EC) fluxes for the FAST were determined after combining its 20 aerosol sizes
into 6 broader size intervals to obtain more total counts for each diameter interval within each 30
min. flux period. These six aerosol sizes covered the range 0.25 µm < diameter < 1.0 µm Thus
aerosol concentrations and fluxes were determined separately from the FAST for the following
diameters: 0.26 µm (0.246 ≤ dia ≤ .28 µm); 0.30 µm (0.28 ≤ dia ≤ .31 µm); 0.33 µm (0.31 ≤ dia ≤
.34 µm); 0.40 µm (0.34 ≤ dia ≤ .44 µm); 0.49 µm (0.44 ≤ dia ≤ .54 µm); and 0.78 µm (0.54 ≤ dia ≤
1.01 µm).. In a 0.1 sec. sampling interval, the FAST typically counted 1 to 10 particles in the larger
diameter (0.5 µm ≤ dia.) intervals but 5 to 40 particles in the for the smaller (e.g., 0.3 µm)
particles. For any 30 minute eddy flux during BEARPEX, the total particle counts in any diameter
interval varied from ~10 4 to 2 x 10 5 .
3.1 Screeni ng, rotations, and mea n removal
3
Periods with at least 28 (of 30) minutes of “valid” data were retained for further analysis. Data were
deemed valid after screening for periods with signal dropouts, people walking on the tower, sonic
anemometer or hygrometer spikes during rain, boom rotation, and poor upwind fetch. To avoid these
problems as well as the result of applying standard micrometeorological screening of eddy flux data
(Vickers and Mahrt, 1997) the data set was reduced to its most ideal periods: 316 total 30-minute
periods with “valid” aerosol eddy fluxes (most were daytime hours) with 67 of these periods also
having aerosol hygroscopic growth measurements (twice each day).
Prevailing nighttime and early morning easterly winds produced EC fluxes that were less
representative because of uneven upwind terrain and advection of emissions from the site’s electrical
generator (located ~125 m. to the N).
After the above data screening steps, tilt and coordinate system rotations were investigated (Lee et
al., 2004; McMillen, 1988) . The derived rotation (“attack”) angle matched the slope of the upwind
terrain such that vertical fluxes were both “surface normal” and streamline normal. This 2 o mean
BEARPEX attack angle derived from the daytime data was similar whether obtained separately for
each 30 min. flux period or alternatively based on mean “attack angle” as a function of wind
direction (Kowalski et al., 1997). These different rotation criteria also produced similar values for
friction velocity (U*) indicating that the fluxes are not sensitive to the rotation criteria.
The eddy correlation heat, vapor, and momentum fluxes that are reported here showed good
agreement with those measured independently by Univ. of California collaborators (Blodgett Forest
Ameriflux site, Goldstein group) for the same time periods (UC= slope * OSU: momentum fluxes:
slope=0.98, r 2 =0.83; sensible heat fluxes, slope=1.15. r2 =0.92; water vapor flux, slope=0.96 r 2 =.77)
despite the fact that their fluxes were measured from a second tower that was located ~10 meters
away (cross wind) and at a lower height agl (13 m for U.C. vs 18.8 m AGL here). This type of
consistency between fluxes measured at different heights and cross wind locations suggests that these
EC measurements are representative of the upwind fetch.
2 Results
4.1 Gen eral
The winds were predictable with upslope winds (S to W) over good fetch from late morning until
sundown each day (S to W) but downslope winds (N to E) at night and early morning. Aerosol
concentrations were quite low and apparently were dominated by biogenic hydrocarbon species
except for a few periods when forest fires elevated particle and gas concentrations. The air was very
dry (15% < RH < 40%) during most daytime periods.
4.2 Eddy correlation flux es
Figure 1 presents the measured diurnal variation of aerosol deposition velocity during BEARPEX.
Aerosol fluxes and deposition velocities were larger during mid-afternoon than mornings or evenings.
Anaysis of daytime data indicates that σw = 1.33 Ustar (r2 =0.93)… demonstrating that the
turbulence at the measurement height is fully developed (Stull, 1988; Foken et al, 2004
4
Figure 1: Diurnal variation of aerosol deposition velocity for six particle diameters as measured
during BEAPREX 2007 (negative is downward). No corrections have been applied to these data. Bars
indicate one data standard deviation for values recorded during that two hour period.
4.3 Counting uncertainti es
The uncertainty for single 30-min. EC aerosol fluxes due to the discrete nature of aerosol “counting”
was calculated as σw/ √N (Fairall, 1984; Nemitz, 19xx; Vong et al., 2003) where N is the number of
particles counted in 30 minutes for a particular particle size and σw is the standard deviation of the
vertical wind velocity. Counting uncertainties, in terms of the particle deposition velocity, (Vd = EC
flux / mean particle concentration) were ± 0.14 to 0.21 cm/sec for the smaller particles (0.25 µm <
dia. < 0.44 µm) but ± 0.55 cm/sec for the larger particles (0.5 µm < dia. < 1.0 µm). The uncertainty
in the pooled values for mean Vd during BEARPEX would be 1/ √ 316, or 6%, of these values for
single 30-min. EC aerosol fluxes. These uncertainties are acceptably small for the smaller (dia < 0.5
5
µm) particles but large enough to reduce confidence in the results for the larger (dia > 0.5 µm)
particles
4.4 Spectra and co-spectra
Figure 2 presents BEARPEX aerosol spectra from the FAST spectrometer, spectra of vertical
velocity, water vapor density ,and virtual temperature, their co-spectra, and the [-5/3] slope that
would be expected for an instrument with an ideal response. While vertical velocity, water vapor,
and virtual temperature behave as expected, aerosol concentration for all six particle diameters does
not. The flattening of particle concentration variance in the spectrum suggests that noise is present
and that the FAST did not fully resolve aerosol concentration above ~0.1 Hz.
Figure 2: Spectra and co-spectra for the indicated variables from 70 daytime periods during
BEARPEX.
There is no appreciable difference between the size aerosol spectra in that flatten out at 2 to 5 x 10-2
The FAST is “noisy” for anything faster than 20 to 50 sec resolution. These spectra for heat, water
vapor and vertical wind speed follow the [–5/3] theoretical slope suggesting that these quantities are
well determined. The third panel in Figure 2 presents the cumulative, normalized covariance for heat,
particles (0.3 µm dia.; other particle diameters are similar), water vapor, and momentum.
6
Although it is clear that the FAST is not an ideal EC scalar sensor based on the co-spectrum for
normalized, (cumulative) fluxes, they are similar (flux occurs over same freq. ranges) to those for
heat, moisture, and momentum for f < 2 x10-1 Hz but the FAST does not capture flux at frequencies
above ca. 5 x 10-1 . Since the heat vapor and momentum fluxes do not have much contribution to the
total flux above this frequency, spectral similarity implies that the FAST only lost a small amount of
the aerosol flux due to its inability to resolve higher frequencies. Most of the aerosol flux is
transported over time scales of 2 to 200 seconds while the heat, vapor and momentum fluxes,
according to the co-spectra, are transported by slightly larger eddies with time scales of 2 to 500
seconds.
The spectra look much better after applying a low-pass noise filter with a cutoff frequency of 0.5 Hz
(half power of the filter) in that they follow the [-5/3] slope in the inertial subrange. Corrections for
the 1 m lateral separation of the aerosol sensor from the sonic anemometer (Moore, 1986) indicate
a loss of flux at higher frequencies of the measured flux but these corrections are not needed here
because the aerosol spectrometer clearly cannot resolve these frequencies but very little of the flux
occurs at these time scales anyway. The scale dependence of the flux is nearly the same for the
unfiltered and filtered time series, indicating that the high-frequency noise does not contribute
significantly to the eddy flux. The effect of averaging time on the computation of eddy covariances
was found to be minimal in that 10- min. or 30-min. mean removal made only 1-2% difference in
the 30 minute fluxes.
4.5 Aerosol hygroscopic growth
The aerosol growth parameter, γ, was calculated twice daily from aerosol size spectra measured by
OPCs operating at two different RH. Equation 1 (Kasten, 1969; Vong et al., 2003; Massling et al.,
2005) relates the measured optical diameters
to yield the hygroscopic growth factor gamma (γ ) as
-γ
D (Shigh ) / D (Slow) = [(1 - Shigh )/(1- Slow)]
eq. 1
where S is saturation ratio (S = RH/ 100%). and D is aerosol optical diameter at the given saturation
ratio (S= RH/100%)
The aerosol were not very hygroscopic. On most days during September 2007 the aerosol were
largely organic (other BEARPEX investigators) and presumably derived from local and regional
biogenic sources.
4.6 Hygroscopic growth correction to aerosol deposition velocity
A correction to aerosol EC fluxes due to any hygroscopic growth of particles can be given as (Vong
et al., 2003; Kowalski, 2001; Fairall, 1984):
ΔVd = - β γ w' S' / (1 –S)
eq. 2
We directly measured all the components of this Vd correction during BEAREPX 2007: the slope fo
the aerosol size distribution (β), the hygroscopic growth factor (γ), the saturation ratio, and the
saturation ratio flux w' S' .
During BEARPEX the hygroscopic growth parameter was measured to be 0 < γ < 0.12 over 258
valid scans with a mean value of γ= 0.05 over the optical size range of 0.3 to 1.0 µm.. There was no
7
difference in measured γ with diameter and no trend with time during BEARPEX. Figure 3 displays
the variation in γ as well as the RH of the scans of the size spectra.
Figure 3: Measured aerosol hygroscopic growth factor (γ) during BEARPEX 2007..
Aerosol hygroscopic growth factor
-! ) for (BEARPEX
This compares to γ = 0.25 for ammonium sulfate aerosol (Vong et al., 2003). The BEARPEX
aerosol was not very hygroscopic compared to sulfate and other combustion-derived particles. From
humidity changes of ‘dry’ (RH ≤ 30%) to 90% RH, the typical BEARPEX aerosol would grow 12%
in diameter while sulfate particles would have grown 78% .
These results are consistent with previous measurements of biogenic/carbon aerosol hygroscopic
growth in the California Sierra (Carrico et al., 2005)
However, the aerosol size distribution during BEARPEX; characterized by the “Junge” slope (β)
[defined from the equation dN/dlogD = c D– β], was calculated for every 30 min. flux interval for each
FAST diameter interval and shown to be “very steep”[ 4 < β < 12 ]. β decreased with increasing
particle diameter. These large values for the Junge slopes indicate that there were lots of smaller
particles to grow into a given diameter interval when RH increased during vertical transport.
Similarly, there were few larger particles that could shrink into a given diameter interval when RH
decreased during vertical transport.
Figure 4: Aerosol size distribution from the OPC during BEARPEX’07
8
The last component of the hygroscopic growth correction to EC aerosol flux is known as the
saturation ratio flux (w’S’, essentially the covariance between RH and vertical velocity; Fairall,
1984).
w' S' = w' q'/ q sat - w' T' (S L v ) / (R v T 2 )
eq.3
During BEARPEX this saturation ratio flux w' S' . was determined from 10 Hz heat and vapor EC
fluxes and was surprisingly large and similar to values observed over grass in Oregon (EFLAT; Vong
et al., 2003 (Figure 5). Positive w' S' were observed at night and negative values during the day.
Figure 5: Frequency of occurrence for saturation ratio fluxes for BEARPEX measurements.
9
During the daytime, measured aerosol deposition velocities change by becoming less downward after a
hygroscopic growth correction. This hygroscopic behavior of the particles reflected the fact that RH
was 1-6% higher at the top of the tower (at 18.8 m AGL) than below (at 7.3 m agl) during daytime
making saturation ratio flux w' S' downward (negative). At night or during the early morning periods
when RH was 5 to 15% higher near the ground than at the top of the tower, measured aerosol
deposition velocities change by becoming more downward after a hygroscopic growth correction.
The hygroscopic growth correction during BEARPEX reduces the magnitude of both upward and
downward aerosol eddy fluxes and deposition velocities compared to observed, uncorrected values.
(Fig 6)
Figure 6 displays the measured (solid line) and hygroscopic growth-corrected (dash line) aerosol
deposition velocities for six different aerosol diameters particles (plotted against fraction velocity
(U*). , as discussed below) In these figures, the plotted uncertainties represent ± one data standard
deviation of the measured values among the 316 total 30-minute observations that are included for
each diameter. Similar results were obtained for particle diameters greater than 0.4 µm.. Although the
hygroscopic growth correction was moderate compared to the earlier EFLAT experiment in Oregon,
it is important to the precise determination of aerosol deposition velocity at any diameter during
BEARPEX.
4.7 WPL correctio ns
Webb-Pearman-Leuning (WPL) corrections, (webb et al., 1980) were performed to determine how
the aerosol deposition velocity was affected by vertical variations in air density associated with T and
RH according to eq. 4
Vd = 1.61 ( w' ! H20 ' /! air ) + (1 + 1.61 q)( w' T' /T)
eq. 4
The Webb correction reached maximum values of 0.05 to 0.15 cm/sec during the middle of each day
(10 am until 3 pm) and was neglible (less than ± 0.02 cm/sec ) in the mornings and evenings. These
corrections are incorporated in the results presented here. However, the aerosol fluxes depended on
moisture variations more through the hygroscpic growth corrections.
4.8 Dependence on frictio n velocity
As shown in Figure 6, aerosol deposition velocity is larger for higher values of friction velocity (U*).
(becomes more negative; negative fluxes and negative Vd imply downward transfer here (towards the
ground), consistent with the sign convention for vertical velocity (w).
10
Figure 6 Aerosol deposition velocities with and without the hygroscopic growth correction.
Aerosol Vd also varied with stability but there were too few stable cases during with good fetch
BEARPEX 2007 to characterize this relationship very well. The maximum Vd occurred during windy
(high U*) afternoon conditions.
4.9 Siz e-dependence of deposition velocity
Figure 7 shows that aerosol Vd during BEARPEX increases with particle size for the smaller
accumulation mode aerosol (0.25 < dia < 0.4 µm); in this figure the plotted errors bars represent the
average counting error for a single 30 minute flux at that diameter. The measured values of Vd for
diameters greater than 0.5 µm are considered less reliable due to noticeably larger counting
uncertainties.
11
Figure 7: Size dependence of aerosol deposition velocity with counting errors (corrected for
hygroscopic growth and WPL).
Size-dependence of aerosol deposition velocity
(corrected for hygroscopic growth) with counting errors,
The uncertainty in the mean values for deposition velocity in Fig.10 are actually 8% (1/ √158) of
those for single 30 minute fluxes but the presentation of counting errors for single 30 min fluxes was
chosen here because it highlights the fact that the results for larger diameters are much less reliable.
5
Discussion we
we we
When we first observed the very low daytime RH at the Blodgett Forest site, we expected that this
hygroscopic growth correction might be small to neglible due to the very low daytime RH and the
organic aerosol composition. However, these hygrophobic particles likely are biogenic and produced
from forest emissions of biogenic hydrocarbons (ref) . These same forest biogenic emissions formed
many smaller particles (dia. < 0.2 µm) through reactions with ozone (ref); these smaller particles
produced in situ resulted in these large values of β. There were so many of these smaller particles
that their modest hygroscopic growth was offset by their relatively abundance. The saturation ratio
flux w' S' . was similar to values reported for EFLAT (Vong et al., 2003) and HAPEX (Kowalski,
2001) during BEARPEX. It is likely that larger vertical velocity fluxuations over the “rough”
Blodgett forest canopy compared to short grass in EFLAT result in better moisture vertical
transport for a given RH gradient.
The apparent decrease in magnitude of aerosol Vd f or diameters greater than 0.4 µm could reflect the
fact that these fluxes are especially noisy due to higher counting errors. It also is plausible that these
larger particles, especially for high U*, no longer follow the motion of turbulent eddies due to
increased particle momentum associated with both larger mass and velocity and thus their
concentration becomes uncorrelated to vertical velocity. We consider that the size dependence of Vd
for dia. ≤ 0.4 µm is correctly characterized here but that values for the larger particles are uncertain.
12
Both the transport of particles by turbulence through the atmospheric surface layer and their
subsequent removal by inertial impaction ought to depend on friction velocity (U*). In addition,
particle inertial impaction ought to depend on diameter because larger particles have more
momentum. Thus, the removal of particles onto a forest canopy and the resulting aerosol Vd ought
to increase with both U* and particle diameter.
These results are similar to those from Gallagher and co-workers from studies that were conducted in
the U.K. in terms of the dependence on U* and the magnitudes (0.3 to 1.0 cm s-1 ) of aerosol Vd .
6
Conclusions
BEARPEX Vd range from 0.2 to 1 cm s-1 and generally agree with previous work by Gallagher and
coworkers. Aerosol deposition clearly increases with both diameter and U* for 0.25 µm < dia < 0.4
µm.
References
BP --organic aerosol references.
Carrico,C.M., S. M.Kreidenweis, W.C.Malm, D.E.Day, T.Lee, J.Carrillo, G.R.McMeeking and
J.L.Collett, Hygroscopic growth behavior of a carbon-dominated aerosol in Yosemite National Park
, Atmos. Environ.39, 1393-1404, 2005.
Fairall,C., Atmos Environ. 1984.
Foken et al., in Handbook of Micrometeorology (Ed. Lee,X., Massman,W., and B.Law), pp181208, Kluwer Academic Press, Dordrecht, 2004
Goldstein, A.H. et al., 2000 site description
Kasten, F., Visibility forecast in the phase of pre-condensation. Tellus, 21, 631-635, 1969.
Kowalski , Atmos Environ. 35, 4843-4851, 2001.
Kowalski et al., J.Atmos Ocean Technol., 14, 468-479, 1997.
Massling,A., M. Stock and A. Wiedensohler, , Diurnal, weekly, and seasonal variation of hygroscopic
properties of submicrometer urban aerosol particles , Atmos. Environ.39, 3911-3922, 2005
Lee,X., Finnigan,J., and K.T. Paw U, in Handbook of Micrometeorology (Ed. Lee,X., Massman,W.,
and B.Law), pp 33-66,, Kluwer Academic Press, Dordrecht, 2004
McMillen, R.T., Boundary Layer Meteorol. 43, 231-245, 1988.
Moore,C.J., Frequency response corrections for eddy correlation systems, Boundary Layer Meteorol.
37, 17-35, 1986.
Nemitz, E. et al., QJRMS 128, 2281-2300, 2002.
Pryor,S. Tellus 60B, 42-75, 2008.
Vickers and Mahrt, J.Atmos Ocean Technol. 14, 512-526., 1997,
13
Vong et al., Tellus 56B, 105-117, 2003
Webb-Pearman-Leuning WPL corrections, QJRMS 106, 85-100, 1980