95-TP58.
EPA/600/A-95/051
Measurements of Dry Deposition
for Deposition Velocity Model Evaluation
Peter L. Finkelstein
John F. Clarke
Atmospheric Sciences Modeling Division
Air Resources Laboratory
National Oceanic and Atmospheric Administration
Research Triangle Park, NC 27711
(On assignment to Atmospheric Research and Exposure Assessment Laboratory
U.S. Environmental Protection Agency)
Thomas G. Ellestad
Atmospheric Characterization and Modeling Division
Atmospheric Research and Exposure Assessment Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
1

-------
95-TP58.04
INTRODUCTION
The U.S. Environmental Protection Agency's (EPA) nationwide network to monitor dry deposition of
gases and particles, the National Dry Deposition Network (NDDN), measures not deposition, but
concentration of pollutants and meteorological variables relevant to deposition processes. The
amount of pollutant being deposited per unit area and time, the flux, is computed as the product of
the measured concentration and calculated deposition velocity. Deposition velocity is estimated using
an inferential model developed by NOAA's Atmospheric Turbulence and Diffusion Division.1'2,3 The
model simulates the physical and chemical processes of pollutant transfer and absorption by plants
and surfaces using measured meteorological and site vegetation variables as input. Annual and
seasonal dry deposition as derived from the inferential model and concentration measurements are
reported for the 50 site EPA NDDN.4
We have recently begun an independent effort to evaluate deposition velocity models by making
direct flux measurements and concurrent meteorological measurements needed for the models at
several NDDN sites, which differ in terrain, climate, soil, and vegetation cover. The measurement
system, instrumentation, and sampling protocol are described briefly herein, along with some
preliminary data from our 1994 field program. More detailed analyses are currently underway.
MEASUREMENT SYSTEMS
The system has three major instrument groups; one to measure trace gas fluxes by eddy correlation or
gradient techniques, one to measure the components of the energy balance, and one to measure the
variables needed by the deposition velocity models; as well as a data acquisition system.
Eddy Correlation Measurements
Fluxes of ozone, sulfur dioxide, carbon dioxide, water vapor, heat, and momentum are measured by
eddy correlation. For the gases, ambient air is pulled from the vicinty of (4 cm) a three axis sonic
anemometer array (= 5m above ground) through a draft tube and filter to the fast response
instruments that are housed in a temperature controlled box at the base of the tower. The fast ozone
analyzer, based on the work of Ray et al.5, employs the chemiluminescent reaction of ozone with
eosin-Y dye, which is borne in a carrier of ethylene glycol. Sulfur dioxide is measured by a fast
response Meloy model SA-285 flame photometric analyzer operated with a direct inlet line and
hydrogen fuel that is spiked with 70 ppb SF6. Fast response data for water vapor and carbon dioxide
are produced by a LICOR model 6262 infrared absorption analyzer.
Draft Tube Considerations. A key feature of our system is the location of the fast response
instruments at the base of the tower in a temperature-controlled box and the use of a draft tube (a 9
m length of Teflon tubing) to draw air from the sonic array to the fast analyzers. We feel that this
design represents the best compromise between locating the instruments on the tower and locating
them in the support trailer some 50 meters away. The advantages of this approach are: 1) no flow
distortion caused by bulky instruments mounted on the tower, 2) minimal sample separation distance
for wind speed, temperature, and gases, 3) lighter and smaller towers may be used, 4) easier
installation and removal of equipment, 5) more accessible gas analyzers for maintenance and
calibration, and 6) corrections for heat flux are not necessary. The disadvantages include: 1) a time
delay between the sampling of the wind velocity and the gas concentration (which can be evaluated),
2

-------
95-TP58.04
2) some loss of high frequency response, and 3) possible draft tube contamination. Locating the
instruments in the support trailer would solve some problems, but would require a large flow rate
through the draft tube, which might distort the flow field in the vicinity of the sonic array. It may
also be untenable for more adsorptive gases such as sulfur dioxide.
Our draft tube is an FEP Teflon™ tube of 0.95 cm inside diameter, 9 m in overall length, with a
Teflon filter located 1.5 m from the tube's inlet, and a flow rate of about 20 L/min. These
conditions result in a Reynolds number of 3000 and a length/radius ratio of 1900, in accord with the
recommendations of Massman6. The Teflon filter is used to keep the tube's inner wall from
becoming contaminated with particles. As a direct test of the high frequency loss caused by our draft
tube, we used our fast response ozone analyzer, which has an inherent response of 5-6 Hz, and
sampled a step change in ozone through various configurations. The analyzer's response degraded to
about 3 Hz, which is corrected as noted below. About half the attenuation is due to the tube and half
to the filter and its holder. Responses for the other analyzers are similarly degraded. The degraded
response, with correction, is still sufficient to perform eddy correlation measurements during most
atmospheric conditions.
To test the response of the system in the field, and derive the correction for high frequency loss, we
examined the co-spectra of the gases with vertical velocity. Figure 1 presents the normalized co-
spectra for four gases and the heat flux, w' T ' taken during mid-day at the Beaufort site. It can be
seen that the energy in the co-spectra (the area under the co-spectral curve) for the gases is indeed
lower than that for the heat flux at the high frequency end of the spectrum, and that it also falls off
more rapidly. This occurs because the frequency response of the gas instruments is not quite as high,
as the temperatuare (1 to 3 Hz vs. 10 Hz); and because some high frequency information is lost due
to mixing in the draft tube. To see that the heat flux spectrum is reasonable we compare the average
of six heat flux spectra taken under good conditions with that proposed by Moore7 based on Kaimal's
Kansas experiment (Figure 2). The agreement at the high frequency end is excellent. An average
gas co-spectra, (averaged over the same six cases and four gases) in the figure shows more clearly
the area of high frequency loss.
To determine whether or not the assumption that the loss is due in part to the draft tube is
reasonable, we consider the transfer function for draft tubes developed by Massman. It is
T(
-------
95-TPS 8.04
which is consistent with that suggested by Moore (op cit), is applied to the computed gas fluxes in
this study.
Data Quality for Eddy Correlation Flux Measurements, In our measurements of the fluxes of 03,
S02, H20, C02, heat, and momentum using eddy correlation techniques we are concerned that the
measurements be as accurate as possible, and not be simply a fortuitous measurement of co-variance
without physical significance. Businger8 presents an excellent review of the theory and practice of
trace-gas flux measurements. In that review he lists a number of possible sources of error which
should be considered in flux measurements. Among them are: 1) concurrent heat and water vapor
fluxes, 2) sampling duration 3) instrument system response time, 4) instrument separation, 5) random
noise, 6) entrainment, advection, and non-stationarity, 7) height of sampling above the surface, 8)
irregular fetch, and 9) flow distortion near the sampling system. We have attempted to take each of
these potential problem areas into consideration, either in the design of the monitoring system, in the
field set-up, or in the analysis and correction of the data. Frequency corrections are discussed above.
We also examined selected spectra for indications of problems.
Spectral analysis of high frequency data collected by the sonic anemometers and fast gas analyzers
can be a good source of information on data quality. Kaimal9 shows idealized spectra (his Figs 2.1
and 2.4) with regions of energy production, dissipation, and the inertial subrange. Analysis of data
by spectral decomposition allow us to observe whether or not problems of advection or non-
stationarity, or problems with fetch, flow distortion, or electronic noise are impacting a particular
sampling period. Advection and non-stationarity will affect the lower frequencies, causing much
higher energy levels than expected in this range. Unusual levels of high frequency turbulence caused
by flow distortion or uneven fetch may show up in the inertial sub-range, while electronic noise will
usually be seen as a continuously increasing energy with increasing frequency in the inertial sub-
range and dissipation ranges. Lack of instrument response will be evident by lack of expected energy
at the high frequency end of the spectrum, assuming there isn't compensating noise.
As a further check on data quality, we examine the net energy balance for each half-hour period, to
be sure that the flux measurements were reasonable. The net flux balance was computed as the
difference of net radiation minus the fluxes of sensible and latent heat, heat fluxes into the ground,
and soil heat storage. If any of the systems is not working properly, it is highly likely that it will be
obvious in this computation. No time period with a net energy imbalance of >150 w/m2 were
included in the final data base.
Gradient System for Nitric Acid Flux
Nitric acid flux is determined from a two-hour measurement of concentration gradient combined
with a vertical exchange coefficient calculated from the temperature gradient and heat flux, similar
to the work of Huebert and Robert10. Concentration measurements are made using filter packs. They
are loaded with Zefluor™ fluorocarbon and Nylasorb™ nylon filters at an analytical laboratory,
shipped to the site, sampled at a flow rate of 15 L min"1, and returned to the laboratory. Analysis of
the nylon filters for nitrate is usually completed within one to two weeks of sampling. In sampling,
the lower units are placed 1-2 m above the vegetation, while the upper units are 7-8 m above the
lower units. For quality control purposes, two filter packs were used at each of the two heights.
Periodically, we ran all four filter packs at one height to check the precision of the overall sampling
and analysis system. During this experiment only one two-hour sample could be collected per day,
4
-------
95-TP58.04
usually during periods of maximum convective activity and other favorable meteorology.
Consequently, the number of valid samples is small.
PRELIMINARY DATA ANALYSIS
Initial results are presented here. While having received scrutiny, these should be considered
preliminary.
Sites
Two sites were studied in the summer and fall of 1994. The first, a coastal site (34.91°N, 76.59°W)
near Beaufort, North Carolina, was on the property of The Open Grounds Farm, a very large,
extremely flat farm which raises corn, soybeans, hay, and cattle. Most of the measurements were
over short grass, but winds sometimes blew over a growing stand of corn. All measurements had an
unobstructed fetch of several kilometers. This site is characterized by strong winds (the sea breeze),
high moisture flux, and low pollutant concentrations. This site was studied during June and July
1994.
The second site was at the Bondville, Illinois, research station (40.05°N, 88.37°W) operated by the
Illinois State Water Survey. This area is also very flat, with excellent fetch. Our measurements were
exclusively over corn. They began in August and extended past the corn's senescence in October.
This site was characterized by moderate winds and moderate pollutant concentrations with occasional
plumes of sulfur dioxide.
Ozone Deposition Velocity
Figure 3, a time series of half-hour fluxes of 03, C02, and latent heat for a six day period at the
Bondville site, is an example of data collected at the two sites. Negative values reflect flux to the
surface. A multiplicative factor of -0.005 for latent heat was used to scale the output for graphical
purposes. Several points are evident from this figure. The daytime fluxes increase rapidly to a
maximum and then decrease just as rapidly, while the nighttime fluxes are relatively constant. The
shape of the three curves is self-consistent for each day, showing the same small scale variation. The
relative magnitude of the peaks of the three gases does not remain constant. Ozone uptake does
occur at night, at about 1/10 the peak value during the day. C02 fluxes are positive at night,
consistent with surface sources. Small scale variations at night are reflected in both the ozone uptake
and carbon dioxide emissions, while the nocturnal water vapor flux is almost zero.
An important goal of this program is to understand the performance of the inferential model of
deposition velocity under various conditions of ambient concentration, meteorology, and vegetation.
A preliminary analysis of model performance is given in Figures 4 and 5 where modeled and
measured 03 deposition velocity, averaged over all days to give an average diurnal cycle, are
compared for both Beaufort and Bondville. Both include only data when the winds were less than 10
m/s. Model results are for the 1993 version of the NOAA multilayer inferential model.11
Several features stand out. Average afternoon measured deposition velocities are slightly higher for
Beaufort than for Bondville, while nighttime values are lower. Modeled afternoon deposition
velocities, however, are nearly twice as high at Beaufort than Bondville. The model underestimates
the afternoon measurements at Bondville and overestimates at Beaufort. The model underestimates
5
-------
95-TP58.04
nighttime deposition of 03 at both sites. We have not yet attempted to account for these differences.
Figure 6 shows the maximum half-hour average ozone flux for each day that had a valid mid-day
set of measurements from August 17 to October 1 at Bondville. These peak values vary considerably
from day to day, depending on the general meteorology of the day. However, it can be seen that
there is a general trend toward smaller peak flux as the season progresses. Also shown is a measure
of the viability of the corn plants, in this case the average number of completely green leaves on
each plant. As corn matures, the leaves brown one by one, first from the bottom up to the ear leaf,
then from the top down to the ear leaf. The variety of corn in this field has a uniform 21 leaves per
plant. It can be seen that the corn plants matured from maximum capability for photosynthesis at the
beginning of the study, to almost none at the end. The reduction in maximum ozone flux seems to
be reflected in the reduced capacity for uptake of ozone by the com as the season progressed.
Nitric Acid Deposition Velocity
For nitric acid, the precision (as the coefficient of variation) achieved for a pair of collocated filter
packs averaged 5.1% at Beaufort and 3.1% at Bondville. The multilayer inferential deposition
velocity model" was run for comparison with measured results (Figure 7). Through the model and
measurements do not show order-of-magnitude differences with each other, there are frequent cases
of differences with each other, there are frequent cases of differences exceeding 25%. In general the
model appears to have more serious overestimates of deposition velocity than underestimates. One
interesting note is the marked difference in the two sites' results. At Beaufort, the model predicts
values in a fairly narrow range of 1 to 2,5 em's0, while the measured values go much lower,
although the lower measured values had increased uncertainly because the gradients were very small
(<5%) in those four cases. In contrast, at Bondville, the measurements ere in a more narrow range
than the model predictions. At this point in our analysis, we do not know the reasons for this
difference. An important part of our work will be examining such cases of poorer agreement, with
the goal of finding patterns of atmospheric or surface conditions when disagreement occurs, which
may lead to improvements in the model or in the measurement network. Most of our runs were
made in the mid-day period when maximum flux and nitric acid concentrations are expected. In
subsequent studies we plan to extend sampling into earlier and later periods and look at diurnal
cycles.
CONCLUSIONS
We have constructed and operated a mobile trace gas flux monitoring system. Data is being
collected over a variety of vegetative covers, land-use characteristics, and climatic regimes. The
complete data set will be used to evaluate and improve deposition velocity models and to study the
mechanisms of gas transfer between the atmosphere and biosphere. Preliminary analysis of the first
year's data shows that the 1993 version of the NOAA multi-layer deposition velocity model is giving
reasonable values, on average, but there are site to site differences, and time of day differences that
need to be understood. These discrepancies tend to confirm the advisability of testing such models
under many different conditions.
ACKNOWLEDGEMENTS
We thank Tilden Meyers of NOAA's Atmospheric Turbulence and Diffusion Division for his
6
-------
95-TP58.04
unselfish assistance in the development of the monitoring system, and supplying us with the
deposition velocity model used in this study. We also thank Jon Bowser, Eric Edgerton, Eric Hebert
and Greg Vetter of Environmental Science & Engineering Inc. who are responsible for the field
operation and logistics of the system; and Alan Fabrick, and Don Reed of A.J.F. Consulting Inc. for
the development of the data acquisition software. Finally we gratefully acknowledge the assistance
of Gabrielle Onorato of Open Grounds Farm; and Don Dolske, Steven Hollinger, and Gary Stensland
of the Illinois State Water Survey for allowing us to use their properties and arranging site logistics.
This paper has been reviewed in accordance with the United States Environmental Protection
Agency's peer and administrative review policies and approved for presentation and publication.
Mention of trade names or commercial products does not constitute endorsement or recommendation
for use.
REFERENCES
1.	B.B. Hicks, R.P. Hosker, Jr., T.P. Meyers and J.D. Womack, "Dry deposition inferential
measurement techniques - I. Design and tests of a prototype meteorological and chemical system for
determining dry deposition," Atmos. Environment. 25A(10):2345-2359 (1991).
2.	T.P. Meyers, B.B. Hicks, R.P. Hosker, Jr., J.D. Womack and L.C. Satterfield, "Dry Deposition
inferential measurement techniques - II. Seasonal and annual deposition rates of sulfur and nitrate,"
Atmos. Environment, 25A(10):2361-2370 (1991).
3.	B.B. Hicks, D.D. Baldocchi, R.P. Hosker, Jr., B.A. Hutchinson, D.R. Matt, R.T. McMillen and
L.C. Satterfield, On the Use of Monitored Air Concentrations to Infer Dry Deposition. NOAA
Technical Memorandum ERL/ARL-141, 1985, 65pp.
4.	J.F. Clarke and E.S. Edgerton, Dry Deposition Flux Calculations for the National Dry Deposition
Network. EPA-600/R - 93/065, U.S. Environmental Protection Agency, Research Triangle Park, NC
27711, 1993, 91pp.
5.	J.D. Ray, D.H. Stedman and G.J. Wendel, "Fast chemiluminescent method for measurement of
ambient ozone," Anal. Chem.. 58:598-600 (1986).
6.	W.J. Massman, "The attenuation of concentration fluctuations in turbulent flow through a tube,"
JGR. 96(D8U 5269-15273 (1991).
7.	C.J. Moore, "Frequency response corrections for eddy correlation systems," Bound.-Laver.
Meteor.. 37:17-35 (1986).
8.	J.A. Businger, "Evaluation of the accuracy with which dry deposition can be measured with
current micrometeorological techniques," J. Clim. and Appl. Meteor.. 25:1100-1124 (1986).
9.	J.C. Kaimal and J.J.Finnigan, Atmospheric Boundary Layer Flows. Their Structure and
Measurement. Oxford University Press, New York, 1994, 290pp.
7
-------
95-TP58.04
10.	B.J. Huebert and C.H. Robert, "The dry deposition of nitric acid to grass," JGR.
90(Dl):2085-2090 (1985).
11.	T.P. Meyers, NOAA ATDD, Oak Ridge TN., Personal Communication, 1994.
8
-------

cr
J
"co
o
O
'u-
J
C/3
c
0.4
0.3
0.2
0.1
0.0
-0.1
I ; I I I
Co-Spectra
Day 167 16:00
wT
a w'Oa'
¦ - w'S02*
~ - - w'HzO'
W'C02*
\
i -1	l—I I l I I > i
- \ 1 I l I
10"3 2 3 4 1 0-2 2 3 4 ] 0-1 2 3 4 1 0° 2 3 4
n
KD
in
Figure 1. Normalized co-spectra, showing a rapid fall off at high frequency for all gases. N is normalized	^
frequency (f-z/u). The abscissa is n times the cospectra normalized by the covariance.	^
03
O
-------
w'T & w'q' Co-spectra
cr
>
o
O
"cr
5'
CO
c
0.2
0.1
0.0
wT
w'q'
M/K
I I )	i	 ¦ 1 ... i i	:	i i . i '	,	1 .i.i.i i	|	| | | | I
TO"3 2 3 4 1 0"2
2 3 4 10"1
n
2 3 4
Figure 2. Co-spectra averaged over 6 cases and all gases, compared with measured heat flux and results from
Moore/Kaimal. An integration under the w' T ' and w'q' curves shows the energy in the gas spectra
to be 17% low due to high frequency losses. The M/K curve is the best fit to the co-spectra presented
by Moore.
-------
FLUX TIME SERIES
BONDVILLE
0.1
-0.1
8-0.2
-0.3
-0.4
-0.5
0.5
-2.5
08/22 08/23 08/24 08/25
DATE
08/26 08/27 08/28
03
C02 H20
Figure 3. Time series of 30 min. average fluxes for 03, C02, and latent heat at Bondville, IL. Some
data are missing. The 03 units are ppb m s"1, the C02 units are ppm m s"\ and the latent
heat units are -0.005 watts m"2.
-------
DEPOSITION VELOCITY DIURNAL CYCLE
Beaufort
-0.2
o
(D
| -0.4
o
-0.6
-0.8
06:00
12:00
00:00
00:00
18:00
•— Measured * Modeled
v	J
Figure 4. Average diurnal cycle of modeled and measured 03 deposition velocity at the Beaufort site.
-------
DEPOSITION VELOCITY DIURNAL CYCLE
Bondville
0
-0.6 -	-	
-0.8 -J	1	1	•	1	1	1	1	
00:00	06:00	12:00	18:00	00:00
Measured * Modeled
V,	„				 ¦'
Figure 5. Average diurnal cycle of modeled and measured 03 deposition velocity at the Bondville site.
-------
MAX 03 FLUX & PLANT VIABILITY
B0NDV1LLE
0
-0.1
1-0.2
x>
CL
Q.
CO
O
-0.3
-0.4
-0.5
IX
08/17 08/27 09/06 09/16
DATE
09/26
20
15 §
Oi
0
10 £
0)
c O
5 Q
0
10/06
x 03
Com
Figure 6. Time series of daily maximum 03 flux and plant viability at Bondville.
-------


HN03 DEPOSITION VELOCITY
MODELED vs OBSERVATIONS



5
•
•
• Bondville
~ Beaufort

In
E
o
4
•
• •



D
LLI
_J
LLI
o
3
• •




•
A •«



O
5
2
1
A A * 9 A \
• A
A




0
(
! . 1 , 1


!

) 1 2 3
OBSERVED (cm/s)

4
5
Figure 7. Measured and modeled deposition velocity for HN03.
-------