PB92-121144
Routine Estimation and Reporting of Dry
Deposition for the U.S.A. Dry Deposition Network
(U.S.) National Oceanic and Atmospheric Administration
Research Triangle Park, NC
Prepared for:
Environmental Protection Agency, Research Triangle Park, NC
1991
[
]

-------
EPA/600/D-91/243
ROUTINE ESTIMATION AND REPORTING OF DRY DEPOSITION FOR THE
U.S.A. DRY DEPOSITION NETWORK
John F. Clarke,* Atmospheric Sciences Modeling Division, Air
Resources Laboratory, NOAA, Research Triangle Park, NC 27711
Eric S. Edgerton, Environmental Science & Engineering, Inc.,
100C Park Forty Plaza, Durham, NC 27713
Rudolph P. Boksleitner, Atmospheric Research and Exposure
Assessment Laboratory, U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
ABSTRACT
A National Dry Deposition Network (NDDN) was established in
the United States during 1986 to document the magnitude,
spatial variability, and trends in dry deposition of ozone
and acidic particles and gases. Currently, the network con-
sists of 50 stations: 41 in the eastern United States and 9
in the western United States. Dry deposition is not measured
directly in the NDDN, but is estimated by an inferential
approach, i.e., fluxes are calculated as the product of meas-
ured ambient concentration and modeled deposition velocity.
The temporal resolution for the dry deposition calculations
is weekly. Chemical species include ozone, sulfate, nitrate,
ammonium, sulfur dioxide and nitric acid. Preliminary dry
deposition calculations yielded the following observations:
(1)	calculated values of dry deposition for colocated sites
are in good agreement suggesting good network precision, and
(2)	spatial patterns of S02 and HN03 dry deposition are
consistent with emission patterns.
1. INTRODUCTION
Lake acidification, forest damage, and other ecological
effects have created interest in quantifying the deposition
of acidic chemical species. Acid species are deposited to
the earth's surface through both wet and dry processes.
Deposition of pollutants by wet processes is relatively easy
to determine through analysis of precipitation samples.
Samples have been routinely analyzed for many sites in North
America since'the late 1970s (NADP, 1990) and spatial
patterns and temporal trends can be characterized from the
wet deposition data. Similar characterizations of dry
deposition patterns and trends have not been made hecause
routine, direct measurement is not practical.
Regional model results (e.g. from the Regional Acid
Deposition Model—RADM) and measurements (see Meyers et
* On assignment to the Atmospheric Research and Exposure
Assessment Laboratory, U. S. Environmental Protection Agency

-------
al., 1990) suggest that dry deposition is a significant
component of the total sulfur and nitrogen deposition budget
and therefore is important to policy and effects issues
including: (1) deposition trends relative to emissions; (2)
pollution effects on vegetation, materials, and ecosystems;
(3) estimation of transboundary flux of pollutants; (4)
patterns and magnitude of total deposition; and (5) evalua-
tion of regional models (pollutant budgets).
Because of the high cost of direct measurement of dry deposi-
tion, a method has been developed to infer dry deposition as
the product of measured concentration and modeled deposition
velocity. The deposition velocity model is based on an under-
standing of the physical and chemical processes of dry depo-
sition as described through measured meteorological and site
parameters. This inferential model (Hicks et al., 1985) has
been used since 1984 in a research network (i.e., the CORE
network) coordinated by the National Oceanic and Atmospheric
Administration (NOAA). Annual and seasonal deposition rates
for total sulfur and total nitrate have been reported for the
nine sites in the CORE network (Meyers et al., 1991).
The U. S. Environmental Protection Agency (EPA) established
the NDDN in 1986 as a monitoring program to characterize dry
deposition. Ambient pollutant concentrations, meteorological
conditions, and land-use data, as required for the inferen-
tial model, are collected at the 50 sites that currently
comprise the network. Recently, a program was initiated to
calculate dry deposition fluxes from these data. This
program and initial results are discussed in this paper.
2. NATIONAL DRY DEPOSITION NETWORK (NDDN)
Currently, the NDDN consists of 41 sites located in the
eastern United States and 9 sites located in the western
United States (see Figure 1).
FIGURE 1. National Dry Deposition Network

-------
At each NDDN site, the following measurements are made: O-j,
S02, SO*2"", NH* + , NO' and HN03, wind speed, wind direction,
standard deviation of wind direction (all at 10 m) , tempera-
ture, relative humidity (at 9 m), solar radiation, precipita-
tion (at 1 m), vertical temperature difference (between 9 m
and 2 m), and surface wetness (approximately 7 to 15 cm).
Meteorological parameters and 03 concentrations are recorded
continuously and reported as hourly averages. The only
exception is the standard deviation of wind direction, which
is reported as the average of four 15-minute averages.
Atmospheric samples for S02, S042", NH4+, N03~, and HN03 are
integrated over a weekly period using a three-stage filter
pack. The filter pack collects particles and selected gases
by passing air at a controlled flow.rate through a sequence
of Teflon, nylon, and base impregnated filters. Filter packs
are changed at each site every Tuesday morning. Blank filter
packs are collected on a monthly basis to evaluate the
passive collection of particles and gases. For additional
information on the NDDN, see Edgerton et al. (1990).
Vegetative cover and land use charactistics influence the
deposition process for S02 and 03 and may also influence
deposition processes for other species. The characteristics
monitored at the NDDN sites include: (1) distribution of
major vegetative species within a 1-km radius, (2) distribu-
tion of other land use types (e.g., urban and water), (3)
Leaf Area Index (LAI) of major vegetation species, and (4)
temporal variation of vegetation activity. The distribution
of land use is obtained from aerial photography and the
vegetation type is obtained by on-site identification. LAI
measurements, using a LI-COR LAI 2000 plant canopy analyzer
(LI-COR, Incorporated, 1989), are made for each of the major
vegetative species at each site at least once each year.
More frequent LAI measurements are made at a few sites to
describe temporal patterns. Each week, site operators record
the status of the surrounding vegetation as percent of summer
maximum leaf area.
The concentration, meteorological, and vegetative data are
combined in the inferential model to calculate the amount of
pollutant that is dry deposited.
3. DRY DEPOSITION CALCULATIONS
3.1 Inferential Model
Assuming that*fluxes are unidirectional (i.e., air to
surface), dry deposition can be expressed as the linear
product of ambient concentration, C, and deposition velocity,
F = C*Vd.	(1)
The influence of meteorological conditions and vegetation are
contained in the deposition velocity, Vd. The inferential
model (Hicks et al., 1985) simulates based on a
theoretical understanding of dry deposition processes. The

-------
procedure is anchored to empirical deposition estimates
obtained primarily by eddy correlation measurements.
The dry deposition process is commonly represented by
analogy to series and parallel resistances in an electrical
circuit. Three major resistance components represent the
physical and chemical processes and each may contain a number
of sub-processes. The total resistance, R, can be written:
R = Ra + Rb + Rc - 1/Vd.	(2)
Ra is the aerodynamic resistance. It is inversely
proportional to the ability of the atmosphere to transfer
material vertically by turbulent processes and is estimated
from easily measured meteorological parameters, i.e., wind
speed and standard deviation of wind direction.
Rjj is the boundary layer resistance to vertical transport
through a shallow (- 1 mm) non-turbulent layer of air in
direct contact with the surface. It depends on aerodynamics
of the surface and physical properties {e.g., diffusivity) of
the pollutant being deposited.
Rc is the surface uptake resistance and accounts for the
direct uptake/absorption of the pollutant by leaves, soil,
and other biological receptors below and within the canopy.
3.2 NDDN Site-Specific Dry Deposition Calculations
Dry deposition is calculated for the NDDN using an updated
version of the Hicks et al. (1935) inferential model. The
meteorological parameters necessary to determine Ra and R^
are obtained from an instrumented 10-m tower at each of the
sites. The surface uptake resistance, Rc, is based on the
observed vegetation characteristics and measured meteorologi-
cal variables {Section 2). The three resistance terms are
calculated hourly and summed to produce hourly values of Vjj
for each chemical species. The hourly values of are then
averaged over a week and multiplied by the weekly integrated
concentrations to produce weekly fluxes of HN03, S042", N03~,
and S02. 0, flux is calculated using hourly measurements of
03 and hourly values of V^.
The NDDN will begin formal periodic reporting of dry deposi-
tion with its 1991 annual report in the spring of 1992. The
report will contain seasonal and annual values for sites that
meet minimum data capture criteria.
3.3 Large-Area Deposition.	- _
Because of potentially large spatial variability of meteoro-
logical conditions, surface features/vegetative cover, and
pollutant concentrations, measured or inferred dry deposition
at a specific site may vary markedly over relatively short
distances. For example, dry deposition to an agricultural
field may be significantly different from that to a nearby

-------
forested area. Because selection of representative sites is
often not practical, an approach has been developed to esti-
mate the bias of a site with respect to a large area sur-
rounding that site.
The approach, documented by McMillen (1990), estimates dry
deposition through the inferential algorithm based on land
use within 40-km to 50-km of the site. Vegetation, surface
roughness, and terrain features are determined on a 1-km
basis using high-resolution satellite data. Meteorological
and concentration data required for the inferertial model are
measured only at a site located near the center of the area;
however, meteorological variability as a function of surface
roughness is incorporated in the approach. Vj and flux are
calculated for each 1-km by 1-km grid cell ana integrated
over the entire area. This approach provides a basis for
estimating potential biases of site-specific dry deposition
calculations and provides large-area deposition for the
evaluation of regional models and for the depiction of
regional dry deposition patterns. The approach, however,
cannot address variability resulting from concentration
variations or mesoscale variability in meteorological
conditions.
4. ACCURACY AND PRECISION OF INFERENTIAL MODEL CALCULATIONS
Because reference methods do not exist for dry deposition,
the accuracy of the inferential dry deposition approach is
difficult to determine empirically and likely will differ
from site to site. A number of comparisons have been made
between the inferential model and dry deposition velocities
obtained by eddy correlation measurements (e.g., see
McMillen, 1990). These studies suggest the uncertainty of
the inferential method for S02 and 03 is probably about 30%
for sites located away from major sources, with uniform
vegetation, and in uncomplicated terrain. Current best
estimates suggest an uncertainty of about 50% for HN03,
S042~, and N03~.
Precision of the deposition estimates by the inferential
approach can be addressed with data from five duplicate NDDN
sites (sites with two sets of identical equipment,
maintenance, calibration, and operating procedures), and
through comparison of five CORE sites (using independent
equipment and procedures) that are colocated with NDDN sites.
Thus, it is possible to develop within-network and between-
network precision estimates. Preliminary results of this
analysis are given below.
4.1 Comparison of NDDN Duplicate Sites
Comparison of data from duplicate NDDN sites focused on depo-
sition velocity. Weekly values of S02 deposition velocity
for duplicate NDDN sites at Alhambra, Illinois (60 km east of
St. Louis) are shown in Figure 2. The plots cover the period
October 1989 through September 1990. The S02 deposition

-------
velocities agree very well for the duplicate sites. Snail
differences (typically < 0.05 cm/sec) are caused by differ-
ences in the measurement of meteorological parameters,
primarily the wind speed and standard deviation of wind
direction.
For the same period, deposition velocities of HN03 at
Alhambra are shown in Figure 3. The agreement is not nearly
as good as for S02• The inferential model for HNOj
deposition is very sensitive to wind speed and to the
standard deviation of wind direction. Small differences in
these measurements on the two towers translate into the large
differences shown in Figure 3.
O
0)
V)
0.6-
0.4-
E
0 0.2
0.0
Alhambra, IL
O NOON Site 157
* NOON Site 257

A.
a
t
$
t
T
f
3
~
A.
FIGURE 2. Weekly S02 deposition velocities for duplicate
NDDN sites at Alhambra, 111., October 1989 - September 1990.
O
!
O
1.6
1.2-
0.8
0.4-
/ 1 A
: 1Y
f ! * . T
\ *
* 1 \ / \
~ i i( \ 1
<* ; \
+* \ / ~
\Ai

u
n
\ *
\
S - V * -
/ ' w	VI v _ «
/' / ^ \
/ «v V	' '
W
^ .
VV
^ —
Alhambra, IL
C NOON S*» 157
• NOONS»»J37
0.0 ^
§ & &

4
FIGURE 3. Same as Figure 2 but for HNO3,

-------
4.2 Comparison of Colocated NDDN and CORE Sites
A comparison of 1989 NDDN and CORE data for West Point, Kev
York, is shown in Figure 4a (S02 deposition velocity) and
Figure 4b (S02 flux). These figures represent between-
network precision, because all measurements (including the
vegetation parameters) and calculations were made by
independent groups. The inferential model, however, is
essentially the saae for both sets of calculations. The
summer differences in Figure 4a are probably caused by
differences in specifying the type and activity of vegetation
as input to the inferential model. The flux calculations in
Figure 4b show larger differences, partly as a result of
differences in the concentration measurements. However, the
general trends for the flux at the two sites are consistent.
0.6
West Point, NY
c CORE SIM
• NOON 6fta
(a)
0.4
o
-------
4
Wert Potnt, NV -
• NOON6R*
0,	,	
///*///
FIGVRE 5. Weekly KNO-j depcsiticr. velocities fcr cclocate;
NDDN and CORE sites at Vest Pcir.t, Now York fcr I5EJ.
Deposition velocities for HNO-j at the West Point cclocated
sites are shown in Figure 5. CORE site deposition velocities
are larger than those for the NDDN site in the summer and
smaller in the winter. The reason for this is not clear;
however, small systematic differences in wind speed and
standard deviation of wind direction have been observed
between the two colocated sites. The inferential model is
very sensitive to these parameters.
Initial analysis suggests that reasonable precision car. be
obtained in the calculation of SO, deposition, but prcfcler.s
may occur in the precision for calculation of HN03 deposi-
tion. Conclusions, however, must await completion cf a
more extensive analysis of all NDDN and CORE duplicate ar.d
colocated sites.
5. NDDN ANNUAL DRY DEPOSITION
Annual average dry deposition velocities and total fluxes fcr
sulfur dioxide for 1990 are shown in Figure 6 fcr the eesterr.
U.S. The values are preliminary as small adjustments tc tr.c
calculation method are likely. In addition uncertainty
bounds, which may vary with site characteristics and terrair
complexity, have only been roughly defined (Section 4) . T.^.e
values represent specific site conditions and are net
necessarily representative of the larger surrounding area.
Regional differences in S02 fluxes are large (primarily due
to regional differences in sulfur dioxide concentration ,
ranging from less than 1 kg/ha in northern New England tc
greater than 10 kg/ha in the Ohio Valley and western Pennsyl-
vania. Regional differences in deposition velocity are less
pronounced, however; values range from 0.14 to 0.31 cm sec.
-------
\»<
A
X*
FIGURE 6. 1990 annual (a) SOj deposition velocities (cm/sec)
and (b) fluxes (kg/ha) for the eastern United States NDDN
sites. (."NS » insufficient data)
A similar analysis for HN03 is shown Figure 7. Nitric acid
fluxes are larger than those for S02, ranging from 2 to 25
kg/ha. High values extend eastward from the Ohio Valley to
the mid-Atlantic region. Smallest values are found in north-
ern New England. Deposition velocities range from about 0.6
to 1.5 cm/sec, primarily reflecting meteorological
variability.
FIGURE 7. Same as Figure 6 but for HN03.
-------
6. SUMMARY AND CONCLUSIONS
Quantitative dry deposition information is needed to charac-
terize spatial patterns and trends in total deposition and to
support ecological effects, modeling, and budget studies.
The fluxes are required for specific sites and also fox
large areas (e.g., 60 km grids). An inferential method for
determining dry deposition as the product of a measured con-
centration and modeled deposition velocity can provide
relevant dry deposition information. The EPA has initiated a
program to calculate reposition fluxes for the NDDN sites and
to assess sources of uncertainty. The program includes
assessment of precision, bi&s, and accuracy through analysis
of duplicated and colocated sites, model sensitivity studies
studies, and direct measurement of fluxes. Formal reporting
of dry deposition fluxes and estimated uncertainties is
expected to begin by spring 1992.
7. REFERENCES
Edgerton E.S., Lavery T.F., Hodges M.G. and Bowser J.J.
(1990) national Dry Deposition Network Second Annual
Progress Report (1988). EPA/600/3-90/020, 88pp.
Hicks B.B., Baldocchi D.D., Hosker R.P., Hutchinson B.A.,
Matt D.R., McMillen R.T., and Satterfield L.C. (1985) On
the use of monitored air concentrations tc inter dry
deposition. NOAA Tech. Memo. ERL ARL-141, 65pp.
LI-COR, Incorporated. (1989) LAI-2000 plant canopy analyzer -
technical information. LI-COR, Inc., 4421 Superior Street,
Lincoln, NE 68504.
Meyers T.P., Hicks B.B., Hosker R.P., Wonack J.D., and
Satterfield L.C. (1991) Dry Deposition Inferential
Measurement Techniques—II. Seasonal and Annual Deposition
Pates of Sulfur and Nitrate. Atmospheric Environment,
25A, 2361-2370.
McMillen R.T. (1990) Estimating the spatial variability of
trace gas deposition velocities. NOAA Tech. Memo. ERL ARL-
181, 37pp.
NADP (1990) NADP/NTN Annual Data Summary, Precipitation
Chemistry in the United States, 1989. Available from the
National Resource Ecology Laboratory, Colorado State
University, Fort Collins, CO 80523.
DISCLAIMER
This document has been reviewed in accordance with U.S.
Environmental Protection Agency policy and approved for
publication. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
-------
DISCUSSION
W.G.N. SLINN. You state that hourly values of Vg are averaged
for a week and then multiplied by the weekly integrated
concentrations to produce weekly fluxes. Do you have some
estimates for the differences between this Vd*c and the more
nearly correct estimates Vd*C. (Similar questions were also
received from S.E. Schwartz and Jan-Willea Erisman.)
J.P. CLARKE. The weekly sampling protocol can result in an
undercalculation of the flux depending on the correlation
between Vg and C. The magnitude of the undercalculation for
O3 and S02 is about 20% in summer and much smaller in winter
(e.g., Meyers, T.P. and Yuen, T.S., 1987, Assessment of
sampling strategies associated with day/night sampling of S02
and O3, J. Geophysical Res., 92, 7605-6712). We have
obtained similar results for NDDN data for hourly 03 and for
a period when S02 and HN03 were sampled on a weekly day/night
protocol. We hope to develop site specific or regional cor-
rection factors to adjust flux calculations for the bias
induced by the weekly sampling protocol.
JAN-WILLEM ERISMAN. It is obvious that the paper summarizes
intentions for interpretation of data in the near future,
with only limited preliminary r*suits. The first result,
however, indicates that the monitoring design and calculation
method is not validated well enough. To ay opinion more
emphasis should be laid to the comparison with the eddy
correlation method and error sources from this comparison.
From experimental work in the Netherlands, it was found that
the method of calculating u* from the standard deviation of
wind direction can better be done directly, rather than
indirectly via estimation of a wind direction dependent ZQ.
The averaging time of the standard deviation of wind
direction also appears to be critical; 5-10 minute averages
should be used rather than hourly averages.
J.F. CLARKE. We plan a program of eddy correlation flux
measurements at selected NDDN sites in 1992-1993 to develop
accuracy and uncertainty information on the inferential
model. We do not have the routine capability to calculate u*
directly at the NDDN sites but will do this along with the
eddy correlation measurements. Standard deviation of wind
direction fluctuations is calculated for 15 minute periods
for the NDDN sites and these values are th6n averaged to
hourly values for application in the inferential model.
JAN-WILLEM ERISMAN. Is surface wetness incorporated in the
Rc estimations?
J.F. CLARKE. Wetness is incorporated into the ^ estimation.
Further, the effect depends on whether the wetness Is the
result of dew formation or due to precipitation. In the
latter case, it is assumed, for S0a, that the water droplets
are saturated as they fall through the atmosphere.
JAN-WILLEM ERISMAN. What is the uncertainty of the
concentration measurements?
-------
J.F. CLARKE. Uncertainty is difficult to define and must take
into account the total sampling system. Acceptance criteria
for 0, is an audit accuracy cf 10%. The accuracy of SO*
and NH4 measurements is also about 10%, S02 is about 15%,
and HN03 about 25%.
WEIGANG GAO. Apparently, the dry deposition flux determined
by an inferential approach is subject to uncertainties in
computing deposition velocities using meteorological
measurements and parameterization of surface resistances.
Your results from the intercomparison show some differences
in Vg between NOAA (CORE network) and EPA (NDDN) systems
during the summer time for species that have a relatively
large surface resistance such as S02, and during all seasons
for species whose deposition is primarily controlled by
turbulent mixing such as HN03 and particles. The uncertainty
of the meteorological measurements should be relatively easy
to remove, but parameterization of surface resistance is more
difficult to include in seasonal change and spatial
variability. Can you clarify what parameterization scheme
and surface parameters were used for inferring surface
resistance in the calculation of deposition velocity? Do you
include LAI data?
J.F. CLARKE. The Inferential algorithms used in the NOAA
(CORE network) and EPA (NDDN) comparisons are essentially the
same; however, the input data were obtained independently.
These data included LAI, percent of area covered by each
vegetation species (NOAA uses two species but the EPA NDDN
allows up to 6), ar.c -cent leaf out. Also, the
meteorological meast nts are measured independently. The
differences in Figur.. Da are due to these parameters. LAI is
measured at all NDDN sites. I believe it is estimated at the
CORE sites.
WEIGANG GAO. The degree to which site-specific deposition
velocity computed for each, station is representative of the
surrounding area is important information needed for
interpretation regional deposition patterns. Has this type
of investigation been conducted for each of the 50 stations?
J.F. CLARKE. Although not implemented at this time,
deposition velocities will be calculated for both the
specific site (1-km radius a ^
-------
TECHNICAL REPORT DATA
{Please read Instructions on the reverse befort complei
1 REPORT NO.
EPA/600/D-91/248
PB92-121144
4* t.tle and subtitle ROUTINE ESTIMATION AND REPORTING
OF DRY DEPOSITION FOR THE U.S.A. DRY DEPOSITION
NETWORK
5 REPORT DATE
6. PERFORMING ORGANIZATION CODE
7. AUTHORIS)
John F. Clarke', Eric Edgerton3 and
Rudolph P. Boksleitner3
8. PERFORMING ORGANIZATION REPORT NO
9. PERFORMING ORGANIZATION NAME AND ADDRESS
1 ASMD/AREAL, RTP, NC 27711
Environmental Science & Engineering, Durham, NC
' AREAL, RTP, NC 27711
10. PROGRAM ELEMENT NO.
N104Q/D/Q6/01 Task 124G(FY-91V
11 CONTRACT/GRANT NO
68-D-80016
12. SPONSORING AGENCY NAME AND ADDRESS
Atmospheric Research and Exposure Assessment Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
13. TyPE of REPORT AND PERIOD COVERED
Preprint (1991)
14 SPONSORING AGENCV CODE
Research
Triangli
NTARVTg
p Park, NP 77711
15. SUPPLEMENTARY^JOTES
16. ABST 1 \CT
A National Dry Deposition Network (NDDN) was established in the United States during 1986 to
document the magnitude, spatial variability, and trends in dry deposition of ozone and acidic particles and
gases. Currently, the network consists of 50 stations: 41 in the eastern United States and 9 in the western
United States. Dry deposition is not measured directly in the NDDN, but is estimated by an inferential
approach, i.e., fluxes are calculated as the product of measured ambient concentration and modeled
deposition velocity. The temporal resolution for the dry deposition calculations is weekly. Chemical
species include ozone, sulfate, nitrate, ammonium, sulfur dioxide and nitric acid. Preliminary dry
deposition calculations yielded the following observations: (1) calculated values of dry deposition for
colocated sites are in good agreement suggesting good network precision, and (2) spatial patterns of SO,
and HNO, dry deposition are consistent with emission patterns.
KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b. IDENTIF IE RS'OPE N ENDED TERMS
c. COSATi 1 iel\l Group



18. DISTRIBUTION STATEMENT
19 SECURITY CLASS (This Report/
UNCLASSIFIED
21. NO. OF PAGES
J RELEASE TO PUBLIC
20 SECURITY CLASSjThtspagei
UNCLASSIFIED
22 PRICE
-------