Sensitivity Analysis of the Multi-Layer Model Used in the Clean Air Status and Trends Network (CASTNET)


4>EPA
United States
Environmental Protection
Agency
   Sensitivity Analysis of the
  Multi-Layer Model Used in the
   Clean Air Status and Trends
      Network (CASTNET)
      RESEARCH AND DEVELOPMENT

-------
EPA/600/R-08/126
October 2008
www.epa.gov
Sensitivity Analysis of the

Multi-Layer Model Used in the

Clean Air Status and Trends

Network (CASTNET)

EPA Contract No. EP-D-05-096

Prepared for

Ralph Baumgardner
U.S. Environmental Protection Agency
Office of Research and Development
National Exposure Research Laboratory
Research Triangle Park, NC 27711

Prepared by

Christopher M. Rogers
MACTEC Engineering & Consulting, Inc.
3901 Carmichael Avenue
Jacksonville, FL 32207

Thomas F. Lavery
Kleinfelder
30 Porter Road
Littleton, MA 01460

Kevin P. Mishoe
MACTEC Engineering & Consulting, Inc.
404 SW 140th Terrace
Newberry, FL 32669
Although this work was reviewed by EPA and approved for publication, it may not necessarily reflect official
Agency policy. Mention of trade names and commercial products does not constitute endorsement or
recommendation for use.
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460 4531 icwa

-------
ABSTRACT
The U.S. Environmental Protection Agency (EPA) established the Clean Air Status and Trends
Network (CASTNET) and its predecessor, the National Dry Deposition Network (NDDN), as
national air quality and meteorological monitoring networks. The purpose of CASTNET is to
track the progress of the 1990 Clean Air Act Amendments (CAAA) emission reduction program
for sulfur dioxide (SOi) and oxides of nitrogen (NOX) in terms of reductions in sulfur and
nitrogen deposition, improved air quality, and changes to affected ecosystems. Both CASTNET
and NDDN were designed to measure concentrations of sulfur and nitrogen gases and particles
and to estimate dry deposition using an inferential model. The design was based on the concept
that atmospheric dry deposition flux could be estimated as Flux = C*Vd, where C represents a
measured air pollutant concentration and Vd represents a modeled deposition velocity. In other
words, the flux is directly proportional to the deposition velocity. Consequently, an uncertainty
in the deposition velocity produces an uncertainly in the flux estimate.

The Multi-Layer Model (MLM), the computer model used to simulate dry deposition, requires
information on meteorological conditions and vegetative cover as model input. Specifically, the
MLM requires hourly averages of wind speed, standard deviation of wind direction (sigma
theta), temperature, relative humidity, and surface wetness. Also as input, the MLM uses plant
speciation data specific to each monitoring site. Speciation data include minimum and maximum
leaf area index (LAI) values and data on the temporal evolution of vegetation leaf-out
characterizing the surroundings of the site within a radius of 1 kilometer (km). Previous studies
have examined the sensitivity of the MLM estimates of deposition velocity and dry deposition
flux to uncertainties in the meteorological and vegetation data and model formulation. For
example, Cooter and Schwede1 concluded that deposition velocity estimates for SO2 and nitric
acid (HNOs) were most sensitive to wind speed and sigma theta. Deposition velocity estimates
for ozone were most sensitive to LAI values. Rogers and his colleagues2 corroborated these
results by showing that annual fluxes of SO2, HNO3, and aerosols were most sensitive to wind
speed and sigma theta.

This research extends the previous sensitivity analyses by investigating the ability to substitute
historical values of deposition velocity or nearby or historical meteorological measurements for
missing on-site data in order to improve the completeness of the dry deposition flux estimates
while not significantly increasing their uncertainty. Using the results from Lavery et al.3, this
study also examines the effect of uncertainties in particulate NCV and FINOs concentrations on
total nitrogen deposition. Finally, this research analyzes the sensitivity of deposition velocity
estimates to the parameterization of non-vegetative surfaces such as rock and water.

INTRODUCTION

The United States Environmental Protection Agency (EPA) established the Clean Air Status and
Trends Network (CASTNET) to provide data for determining relationships between changes in
emissions and any subsequent changes.in air quality, atmospheric deposition, and ecological
effects. The rural monitoring network was mandated by the 1990 Clean Air Act Amendments

-------
(CAAA) to assess the effectiveness of requirements promulgated to reduce emissions of SC>2 and
NOX as Congress recognized the need to track real-world environmental results as the Acid Rain
Program was implemented.
X
Under the CAAA, the Acid Rain Program has produced significant reductions in SOa and NOX
emissions from electric generating plants since 1995. More recent NOX emission control
programs also produced substantive declines in NOX emissions in the eastern United States.
These programs include the Ozone Transport Commission (OTC) NOX Budget, NOX State
Implementation Plan (SIP) Call, and NOX Budget Trading Program. EPA relies on CASTNET
and other long-term monitoring networks to generate the data and information used to assess the
effectiveness of these emission control programs under several different mandates including the
Government Performance and Results Act, The National Acid Precipitation Assessment Program
(NAPAP), Title IX of the CAAA, and the United States - Canada Air Quality Agreement.

CASTNET has its origins with the National Dry Deposition Network (NDDN), which was
established in 1986 and began operation in 1987. Many of the original NDDN sites are still
operational after 20 years and provide useful information on trends in air quality.

CASTNET was designed primarily to measure seasonal and annual average concentrations
and depositions over many years. Consequently, measurements of weekly average
concentrations were selected as the basic sampling strategy. An open-face, three-stage filter
pack was employed to measure gaseous and particulate sulfur and nitrogen pollutants as well
as concentrations of other pollutant species. The filter pack technology and sampling
protocol have been used consistently over the 20 years, providing a comparable data set each
year and allowing for the analysis of long-term trends.

CASTNET is sponsored by EPA and the National Park Service (NFS). NPS began its
participation in CASTNET in 1994 under an agreement with EPA. NPS is responsible for the
protection and enhancement of air quality and related values in national parks and wilderness
areas. The CASTNET sites sponsored by NPS as of September 2008 numbered 25. At that time,
the network included 84 monitoring stations at 82 site locations throughout the continental
United States, Alaska, and Canada, as shown in Figure 1 . For more information on CASTNET,
please visit http://www.epa.gov/castnet/4.

-------
Figure 1. Locations of CASTNET Sites as of September 2008
Current Sites:
• EPA Sponsored
• NPS Sponsored
- Collocated Pairs
Dry deposition processes are modeled as resistances to deposition. The original network design
was based on the assumption that dry deposition or flux could be estimated as the linear product
of measured pollutant concentration (C) and modeled deposition velocity (V
-------
The influence of meteorological conditions and vegetation on the dry deposition process is
simulated by Vd. Deposition velocity is calculated using an inferential model that uses
resistances to deposition to model the naturally occurring dry deposition processes5'6'7'8.

Equation 2. Estimation of deposition velocity.
vd = (Ra + Rb +RC)-

The schematic of the MLM in Figure 2 shows the relationships among the various resistances
and illustrates the meteorological and other data that are required as model input. An improved
version of the MLM was produced by EPA's Office of Research and Development (ORD)
during 20069. This version includes changes to the soil moisture factor, which affects the
stomatal and soil resistances, and to the radiation algorithm, which also affects the stomatal
resistance. All deposition velocities and fluxes for the entire network were recalculated using the
updated model.
Figure 2. Multi-Layer Model
Flux = C X V d
1 /V —
1/Vd ' 1 1
I _ 1 _

1
+ Y
• a
'a, soil
r + r, v + r,
• s ' b 'cut ' b

= turbulence
= turbulence near soil
= thin layer at surface
'a,soil 'soil
'soil
a, soil

soil-
Wetness,
Precip
-------
As shown in Figure 3, CASTNET related measurements and estimates of the dry deposition
component of total deposition are being used to demonstrate the successes and limitations of
CAAA programs.

Figure 3. Time Series of Dry Sulfur Deposition (kg/ha/yr) from 1990 through 2007
12
10 -
^
ro 8
^
15)
^ 6
o
& 4^
90th Percentile
75th Percentile
Median
Mean
25th Percentile
10m Percentile

-+-
T- CM CO
O O) O>
O) o) a)
-
O5
O)
oooooooo
CXICMCMCMCMCMCMCM
DATA

Data collected from CASTNET sites fall into two general categories: continuous and discrete.
Continuous data include measurements obtained from the field program by instrumentation
located at each site. They include meteorological parameters, ozone, and flow through the filter
pack. Discrete data consist of laboratory results acquired through analyses of the 3-stage filter
packs installed weekly at each site. These data products are combined by the MLM to create
estimates of dry deposition, which is the end goal of the network. All data described are
processed and validated, and instrumentation is calibrated, according to protocol described in the
CASTNET Quality Assurance Project Plan (QAPP)10.

Continuous Field Measurements

Continuous data are polled from each site via telephone or cellular modem. The on-site data
acquisition system (DAS) calculates hourly averages for the meteorological parameters measured
which are transferred to the Data Management Center.
• temperature (°C)
• difference in temperature between 2 and 9 meters (°C)
• solar radiation (W/m2)
• relative humidity (%)
-------
• precipitation (mm)
• scalar wind speed (m/s)
• vector wind speed (m/s)
• wind direction (degrees)
• standard deviation of the wind direction within the hour
• wetness
• ozone (ppb)
• rate of flow through the filter pack (liters/min)

Of these, the MLM uses wind speed, standard deviation of the wind direction, solar radiation,
temperature, relative humidity, precipitation, and wetness to parameterize soil moisture and to
calculate estimates of hourly deposition velocity.

Dry Deposition Filter Pack Concentration Data

The filter pack used at dry deposition configured CASTNET sites consists of three types of
filters stacked in the following order: Teflon®, nylon, and cellulose. Filter packs are installed
and run for a seven-day period. The Teflon® filter removes particulate SO42", NOs", NH4+, Ca2+,
Mg+, Na+, and K+. The nylon filter removes HNOs and some 862. Finally, two cellulose filters,
which consist of a cellulose fiber base impregnated with potassium carbonate (K^COs), remove
SC>2. The nylon filter SC>2 and cellulose filters SC>2 are summed to provide weekly total 862
concentrations. The nylon filter HNOs is converted to NOs" and added to the NOs" collected on
the Teflon® filter to provide weekly total NCV concentrations. Final atmospheric concentrations
are calculated by multiplying the analyte concentrations (in micrograms) with the aggregated
flow volume (in meters cubed) for the time period the filter pack was installed. These
concentrations are maintained as micrograms per meter cubed.

Plant Speciation Information

The MLM also requires data regarding the characteristics of the plant types that immediately
surround the site. The CASTNET database includes this information for all of the vegetation in
a 1 km radius around each site. The data include plant type and percent coverage within the
1 km radius. For each plant species, data include leaf area index minimum and maximum
values, canopy profile, a static annual percent leaf out distribution, canopy height, and minimum
stomatal resistance. Also, information regarding preferred temperature ranges for a plant type is
included.

Dry Deposition Estimates from the MLM

The MLM simulates hourly average deposition velocities in centimeters per second (cm/s) for
SC>2, HNOs, Oa, and particles (SC>42~, N(V, and NH4+). The estimates are dependent on the
specific meteorological measurements and plant type information described previously. The
hourly deposition velocities are combined with the filter pack and 63 concentrations to estimate
-------
hourly rates of dry deposition, in kilograms per hectare (kg/ha). The MLM post-processor
aggregates the hourly deposition velocity and deposition rates into weekly, monthly, quarterly,
seasonal, and annual values. Each aggregation step requires a percent completeness of
69 percent for the relevant underlying values. For example, 69 percent of the hourly values
within a week must be valid in order for the weekly value to be valid. Weekly and monthly
values are calculated from hourly values. Seasonal and quarterly values are calculated from
weekly values. Annual values may be calculated from either seasonal or quarterly values, but
standard CASTNET protocol is to calculate annual aggregates from the four quarterly values
within a calendar year. Table 1 shows the aggregation criteria.
Table 1. Aggregation Percent Completeness Criteria
Aggregation Period
Weekly
Monthly
Seasonal
Quarterly
Annual
Requirement * '
69% of hourly values
must be valid.
69% of hourly values
must be valid.
69% of weekly values
must be valid.
69% of weekly values
must be valid.
At least 3 out of 4 quarters
(or seasons) must be valid.

METHODS

Because of the percent completeness requirements discussed in the DATA section, annual
deposition estimates for many sites are unavailable in large part because of invalid
meteorological inputs. In order to get a valid hourly deposition velocity estimate, wind speed,
standard deviation of the wind direction, solar radiation, temperature, relative humidity, and
precipitation values must all be valid. The failure of a single meteorological parameter during a
semi-annual calibration visit to a site can lead to the loss of a valid flux estimate for the year.
Because of the reliability of the flow system, concentration data are usually available and only
occasionally cause the loss of an annual estimate. Table 2 shows the percentage of site-year
estimates for 1987-2007 (all available) that are valid with a breakdown of how many are
calculated from four valid quarters as opposed to three valid quarters. Approximately 1 in 4 site-
years are not available because of incomplete data. Because of this, it is desirable to develop a
protocol for replacing the missing deposition velocity data using a summary or historical value.
-------
Table 2. Summary of Valid Site-Years

total
4 quarters valid
3 quarters valid
All valid site-years
(3 or 4 quarters valid)
- '$Np««K
1404
711
291
1002
•' •'**«**'•
Completeness

51%
21%
71%
This study is divided into five tasks (A-E) that address the requirements of Task 2 from the Work
Plan.
A) Calculate 10-year mean deposition velocities (1998-2007) for all sites. Compare 10-year Vj
values from all sites against one another to identify potential "near site" pairs. Compare sites
against themselves to determine representativeness of 10-year mean. Analyze results with
respect to current list of "near sites" used in MLM runs.

B) Using estimates of the uncertainties of NOs" and HNOs concentrations from Baumgardner et
al. (2008), review the effects of these uncertainties on the weekly and annual dry deposition
fluxes of NOs", HNOs, and, most importantly, total measured nitrogen deposition
(NO3-+HNO3+NH4+, as N).

C) Estimate annual fluxes by using the 10-year mean deposition velocities calculated during
Task A along with annual mean concentrations. Compare the fluxes calculated using the 10-year
mean deposition velocities with the actual (historical) annual flux estimates and with annual flux
estimates calculated from annual mean deposition velocities and annual mean concentrations.
Determine the differences that arise from using site-year with either three or four valid quarters
(all valid site-years) compared with using only site-years with four valid quarters (complete year
is represented).

D) Examine the effect on deposition velocities from modifying the parameterization of non-
vegetative plant types such as rock and water.

E) Estimate deposition velocities and fluxes by replacing missing vector wind speed and sigma
theta (standard deviation of the wind direction) data with near site values or historical weekly
mean values. Compare these estimates with the actual annual flux estimates.

Task A:
Ten-year mean deposition velocities for 1998-2007 were calculated for each site. Each site was
compared against every other site regardless of geographic proximity. All site combinations
where the percent difference between each of the estimated deposition velocities (SOi, HNOs,
and particles) was ±10 percent were identified. This list was compared against the current list of
"near sites" used for replacing missing meteorological data when conducting a MLM model run.
-------
Each site was also compared against itself by comparing the 10-year mean deposition velocity
with each of the ten annual mean depositing velocities from 1998-2007.

Task B:
The uncertainties of NCV and HNOs concentrations were estimated using results of the
Maryland Aerosol Research and Characterization (MARCH) / CASTNET comparison from
Lavery et al.3. MARCH was a state-of-the-science monitoring project conducted at Ft. Meade,
MD. Results from the MARCH study were compared with results from the CASTNET site at
Beltsville, MD (BEL 116), which is approximately 10 km southwest of Ft. Meade. As previously
described, CASTNET uses an open-face, 3-stage filter pack (Figure 4) to collect particles and
gases. Particulate NCV is collected on the Teflon® filter (the first filter) and HNOa is collected
on the nylon (second) filter. There are several issues with this collection protocol with respect to
these two analytes. First, in warm conditions, particulate NCV may volatilize off the Teflon®
filter and be captured by the nylon filter. When analyzed, the recaptured NCV will appear to be
HNCV thus artificially increasing the HNOs concentration for the week. Second, because the
filter pack hangs on the tower for a week, HNOs may be released after capture on the nylon filter
leading to lower HNOs concentrations. Finally, because the CASTNET filter pack is an open-
face measurement device, it captures both large and small NCV particles. There is evidence that
a significant portion of the CASTNET measured NCV is actually sodium nitrate (NaNOs) and
calcium nitrate [CafNOs^] with a smaller portion attributable to ammonium nitrate (N^NOs).
This is a problem for deposition flux estimates because the particulate deposition velocity used to
calculate fluxes of NCV is formulated using the assumption that the particle being deposited is
small (less than 2.5 um) such as NILtNOs or ammonium sulfate [(NH^SCU]. The deposition
velocity for larger particles (greater than 5 |im) such as NaNOs and Ca(NO3)2 would be higher
due in part to gravitational settling, which is not considered by the MLM.

Figure 4. Filter Pack Schematic
Quick Disconnect
Shipping Cap
(removed during sampling)
Teflon* Spacers
10
-------
Despite the volatilization of NOs" off the Teflon filter and subsequent collection on the nylon
filter, which appears to be HNOs, the HNOa concentrations at BEL1 16 were still approximately
15 percent lower than the MARCH HNOs concentrations. The deposition velocity for HNOs is
an order of magnitude higher than that for MV and NH4+, so any increase in HNOs
concentrations would have a dramatic effect on total estimated nitrogen deposition. For NCV,
the comparison with MARCH led to a conclusion that, if considering only N^NOa (or more
generally, NXV particles smaller than 2.5 um), CASTNET NO3 concentration may be as much
as 200 to 400 percent higher than actual N^NOs concentrations especially during warmer
months.
The effects of these uncertainties on the weekly and annual dry deposition fluxes of NCV
and, most importantly, total measured nitrogen deposition (NCV+HNO3+NH4+, as N) were
reviewed by manipulating the annual flux values for BEL1 16 by modifying the percentages with
the percentages listed above and recalculating the total measured nitrogen value.

Task C:
Task C involves making comparisons between archived, historical MLM simulations of annual
deposition rates with annual deposition estimates based on historical deposition velocity data. In
particular, annual MLM results for each often years from the period 1998 through 2007 were
compared with flux estimates calculated in two ways. The first involved multiplying 10-year
mean deposition velocity (calculated during the performance of Task A) for each site for each
pollutant by the measured annual mean concentration for each pollutant and comparing with the
MLM results. The second involved multiplying annual mean deposition velocities for each year
by the measured annual mean concentrations and then comparing with the respective MLM
simulations.

Task D:
There are five non-vegetative plant types used by the MLM:
o WATER,
o ROCK,
o SAND,
o URBAN, and
o LAVA.

There is a sixth, MIXED RURAL, that is partly non-vegetative and partly vegetative. In July
2002 as part of an update of the plant related database that supports MLM operation, the
parameterizations of WATER and ROCK were modified. To examine any changes this change
introduced, model runs were conducted for two sites, Chiricahua National Monument, AZ
(CHA467) and Indian River Lagoon, FL (IRL141), using the original parameterization and the
updated parameterization. Differences in the model run results were evaluated.
11
-------
Although the Work Plan called for manipulating the Leaf Area Index (LAI) values for the non-
vegetative species, this does not make sense to do from a physical standpoint. All of the LAI
values for all non-vegetative plant types have always been 0.0. Therefore, no model runs were
conducted with manipulated LAI values.

Task E:
Missing vector wind speed and sigma theta (standard deviation of the wind direction) data for the
site at Lykens, OH (LYK123) were replaced with data from the Deer Creek, OH site (DCP114),
which was determined from the analysis in Task A, and with historical weekly mean values from
LYK123. These two parameter have been shown by previous sensitivity analyses to be the most
influential in the estimation of deposition velocities1'2. The periods of missing data were
artificially created using a scheme described below. Deposition velocity and flux estimates were
calculated for 2007 by running the MLM using the replaced meteorological data. In total, 48
"sites" were run through the model. Each site represented a different combination of missing
and replacement data. These estimates were compared with the actual annual flux estimates to
determine if near site data or historical average values are better sources for substitution of
missing data.

To create the dataset with missing data, values were removed from a specific number of
randomly selected weeks for several parameters. Different substitution values could then be
easily adapted to the same missing data scheme. Data were removed from six levels of
completeness in two different quarters. First and third quarters were selected because they
represent the weather extremes of winter and summer seasons. A site identification code was
created to represent each combination of missing data. The result was a Cartesian product of
each grouping: 12 identification codes per parameter, per site. Week numbers were randomly
selected for each group of missing weeks and the same weeks were used for each site in the
group. For example, weeks 1 and 2 were randomly selected for the Two Missing Weeks group
and every site and parameter combination used these weeks for the Two Missing Weeks set. The
value for the identified parameter was set to Null for each data point within the selected weeks.

Figure 5 shows the construction of the site id used to document the periods of missing data.
Table 3 describes the meaning of the X, Y, and Z positions of the site id. Table 4 documents the
weeks that were randomly removed for each scenario. Sites where missing data were replaced
with data from DCP114 begin with a LYK site identifier. Sites where missing data were
replaced with historical weekly averages begin with a HIS site identifier. The historical weekly
averages were obtained by calculating the average of all valid sigma theta or vector wind speed
for a particular week for the 10-year period of 1998-2007. This created a data set of 54 weekly
averages for these two parameters. Missing hourly data were replaced by determining the week
of the missing hour and then substituting the average value for that specific week.
12
-------
Figure 5. Site Identification Codes
AAAXYZ
Three letter Site Identifier-)
Quarter Identifier-
Number of Missing Weeks Identifier-
Substituted Parameter Identifier—
Table 3. Description of X, Y, and Z Codin
O -> Quarter One
T -> Quarter Three
T -> Two Weeks
F -> Four Weeks
S -> Six Weeks
E -> Eight Weeks
N -> Ten Weeks
A -> All Weeks
S -> Windspeed
G -> Sigma Theta
Table 4. Randomly Selected Weeks Removed
Two
Four
Six
Eight
Ten
All
1,2
2,3,8,9
5,6,9,10,11,12
1,2,4,5,7,8,9,10
1,2,3,4,5,6,7,8,9,12
All
Examples:
o LYKOFS - Four weeks of windspeed data were removed from quarter one at site
LYK123 and replaced with data from DCP114.
o HISTAG - All weeks of sigma theta data were removed from quarter three at site
LYK123 and replaced with historical weekly averages.

Several important notes on these model runs:
o No time was provided for spin up of the soil moisture parameterization. All runs were
initialized with a soil moisture values of 20 mm. Because all runs started with the same
value, the spin up is not important to the conclusions of the study.
o Scalar wind speed was not used in the model runs. In normal model runs, it can be used
as a backup when vector wind speed is either invalid or missing.
13
-------
RESULTS

Task A:
The first analysis performed for Task A was a comparison of the 10-year mean deposition
velocities for 1998-2007 for each site with the 10-year mean values for all other sites. All
analyses for this tasks used site-years where all four quarters used in the annual mean were valid
to ensure that all seasons were represented. The purpose is to identify and pairs of sites where
the 10-year mean deposition velocities are similar thereby offering an indication that the sites
could be used as "near sites" for the purposes of replacing missing meteorological data. For
reference, the currently assigned "near sites" are presented in Table 5. These "near sites" have
been in use throughout the history of the network. There are up to two "near sites" for each site.

The current "near sites" were assigned based on geographic proximity and similar terrain and
land use. They were not assigned and had not been reviewed using any type of quantitative
analysis to determine their value as a replacement data source. Also, it is possible that these
assignments may have considered the comparability of the concentration values between the
sites, which bears no relevance on the utility of the meteorological data as a source of data for
replacing missing data at a nearby site.

Relative percent differences (RPD) were calculated for 862 Vd, HNOs Va, and particulate Vd.
Comparisons were examined to determine which site pairs had a percent different for each of the
three deposition velocities of ±10 percent. The results are shown in Table 6. Rows that are color
coded blue show sites that are currently designated as "near sites".

Rows that are color coded orange show the RPD values for the two pairs of collocated sites. The
single row highlighted yellow designates the "near site" pair that was selected for use in the Task
E section of this study (LYK123 and DCP114, which are currently designated as "near sites").
The N/S (North/South) Distance and E/W (East/West) Distance columns show the distance in
kilometers that separate the sites.

Interestingly, only 5 current "near site" pairs are identified through this analysis as being a good
match for replacing missing meteorological data. Most of the matches shown in this table are
between sites that are not near one another. Therefore, they cannot be used as near sites and the
good comparison is not related to the influence of similar synoptic scale meteorological
conditions.

Also of interest is the comparison between the collocated sites. These pairs are located at the
same monitoring location and should have the best comparisons. However, the RPD for SO2 Vd
is approximately 5 percent for MCK131/231 and -5 percent for ROM206/406. The RPD for
Va and particulate Vd are closer to zero.
14
-------
Table 5. Current Assigned Near Sites for All CASTNET Sites
CAsmrrtatJif ::•"•
ABT147
ACA416
ALC188
ALH157
ALH257
ANA115
ANL146
ARE128
ARE228
ASH 135
ASH235
BBE401
BEL116
BFT142
BVL130
BWR139
CAD150
CAN407
CAT175
CDR119
CDZ171
CHA267
CHA467
CHE185
CKT136
CND125
CNT169
CON 186
COW137
COW182
CTH110
CVL151
DCP114
DCP214
DEN417
DEV412
EGB181
EGB281
ESP127
EVE419
GAS153
GAS253
GLR468
GRB411
GRC474
GRS420
GTH161
HBR183
HOW132
HOX148
HVT424
HWF187
IRL141
JOT403
KEF112
KNZ184
KVA428
LAV410
nearsitet

VIN140
VIN140
LYK123
BVL130
BEL116
BEL116
WST109
WST109

ARE128

ALH157

CVL151

PAR107

PED108

SPD111

KEF112
CAD150
LYK123
LYK123

MCK131

SND152
SND152

PRK134

CAT175

CTH110

nearstte2

BVL130
BVL130
SAL133
SAL133
PSU106
PSU106

WSP144

SAL133

LRL117

LCW121

MKG113
SND152
OXF122
OXF122

SPD111

ANA115

LYE 145

MKG113

"
f

•*f* *^

t ' j"
^^

1 V/Y
71- v
L i T * V

4 1-'

•^

~°t --
^

' ,^ <
•

-i "i
1

t,
A,- '
-,

_, / '
^ ^

- : '

i- #«\ a

laBref^tw:": ,
LCW121
LRL117
LYE145
LYK123
MAC426
MCK131
MCK231
MEV405
MKG113
MOR409
NCS415
OLY421
ONL102
OXF122
PAL190
PAR107
PAR207
PBF129
PED108
PET427
PIN414
PND165
PNF126
POF425
PRK134
PSU106
QAK172
RCK163
RCK263
ROM206
ROM406
RTP101
SAL133
SAN 189
SAV164
SCR180
SEK402
SEK430
SHN418
SND152
SPD111
STK138
SUM156
SUM256
THR422
UIN162
UVL124
VII423
VIN140
VOY413
VPI120
WEL149
WFM105
WNC429
WPA103
WPB104
WSP144
WST109
.nMR&Kl
SPD111
PAR107

DCP114

ESP127
ESP127

PSU106

ESP127
DCP114

LRL117
LRL117
ESP127
RTP101

VPI120

WEL149
ARE 128

PED108
BVL130

ARE128
GAS153
ESP127

WEL149

ALH157

SHN118
PRK134
WST109

WPB104
WPA103
BEL116
WFM105
nearsite2
CDR119
KEF112

OXF122

SPD111
SPD111

KEF112

PBF129
LYK123

CDR119
CDR119
SPD111

ESP127

MKG113

LYK123

VPI120
ESP127
PBF129

ANA115

BVL130

PNF126
ANA115

WSP144
WSP144
ARE128

15
-------
Table 6. All Site Pairs with a Percent Difference of ±10 Percent for each Deposition Velocity
site id (1)
ABT147
ABT147
ALC188
ALC188
ALC188
ALC188
ALC188
ALC188
ALH157*
ANA115
ANA115
ANA115
ANA115
ARE128
ARE128
ARE 128
ARE 128
ARE 128
ASH 135
ASH235
BEL116
BEL116
BEL116
BFT142
BFT142
BFT142
BWR139
BWR139
BWR139
CAD 150
CAN407
CDZ171
CDZ171
CDZ171
CDZ171
CDZ171
CDZ171
CHE185
CHE185
CHE185
CKT136
CND125
CND125
CON186
CTH110
CVL151
CVL151
CVL151
DCP114*
DCP114
EGB181
EGB181
EVE419
GAS 153
GLR468
HOW132
HOX148
HOX148*
HOX148
HVT424
LAV410
LYE 145
LYK123*
site id (2)
CHE185
SHN418
ARE 128
ASH235
DCP114
GAS 153
LYK123
SND152
BVL130*
BWR139
EGB181
HOW132
VPI120
CKT136
GAS 153
LYK123
OXF122
WSP144*
CHE185
DCP114
CVL151
EGB181
PRK134
PNF126
QAK172
SND152
HOW 132
SAL133
VPI120
KEF112
JOT403
DCP114
MCK231
PAR107
SND152
STK138
VIN140
CKT136
LYE 145
SHN418
SHN418
HOW132
PND165
MEV405
OXF122
EGB181
PRK134
SUM156
LYK123*
SND152
PRK134
SUM156
PND165
LYK123
MOR409
VPI120
PED108
PRK134
SUM156
SAN 189
PIN414
VPI120
OXF122*
N/S distance (km)
676
368
-1055
-1796
-1023
-306
-1165
-429
-131
441
-202
-311
565
222
749
-110
43
-43
1205
773
558
-578
-686
-136
-561
66
-752
-263
124
-823
487
-317
-102
-256
277
-611
-217
-241
-810
-308
-67
-1105
-851
-333
318
-1135
-1244
432
-142
594
-108
1568
-1947
-859
194
875
779
-114
1562
-2598
450
635
154
E/W distance (km)
1813
514
-1368
-2079
-892
-800
-913
-675
-100
-623
-330
-1215
-268
461
568
455
594
-195
2101
1188
1039
237
1102
434
377
748
-592
764
356
-1147
526
-367
-224
-655
-150
172
-29
-928
-1729
-1299
-371
-890
2396
-674
646
-801
64
-385
-21
217
865
417
2329
-113
650
948
-595
389
-60
-4591
-34
600
138
so2 vd RPD
-7.7
-8.8
6.9
0.6
-4.5
5.7
-0.2
-5.0
2.1
-3.9
8.8
3.2
-0.3
9.7
-1.2
-7.0
-0.5
-4.3
9.3
-5.2
-3.2
-8.4
-5.8
-0.1
-9.3
5.3
7.1
-8.5
3.7
-0.7
-4.6
6.7
-8.3
9.5
6.2
-5.9
2.5
-5.2
6.1
-1.2
4.1
-2.8
7.8
6.1
-7.9
-5.2
-2.7
-2.3
4.4
-0.5
2.6
3.0
-5.1
-5.8
0.2
-3.5
6.0
-7.7
-7.3
-4.6
-5.8
7.6
6.5
hno3 vd RPD
-8.3
-3.3
6.7
-8.9
-2.4
-0.2
3.5
6.9
-2.6
1.1
6.9
-6.9
-4.6
-7.4
-6.8
-3.2
5.1
4.4
2.5
6.5
-4.0
1.4
2.2
-5.1
-9.6
9.7
-8.0
9.8
-5.7
2.3
-1.4
-8.9
-4.1
4.5
0.5
-5.9
3.6
3.7
5.8
5.0
1.3
-6.6
-9.0
0.1
1.6
5.4
6.2
9.5
5.9
9.4
0.8
4.1
-4.1
3.6
-0.9
2.3
9.1
-3.9
-0.6
-1.7
-0.3
6.3
8.4
participate vd RPD
9.9
1.8
-3.5
4.1
0.8
-7.8
-0.2
-2.0
-5.3
1.0
-0.8
-2.2
1.6
-5.7
-4.3
3.3
4.8
5.2
-7.0
-3.3
7.6
4.6
7.2
-7.1
9.7
8.9
-3.2
6.9
0.6
0.9
7.3
1.6
6.7
8.3
-1.2
7.3
6.1
-6.5
2.0
-8.1
-1.6
8.7
-2.7
-2.8
-8.9
-3.0
-0.4
5.1
-1.0
-2.8
2.6
8.1
3.7
7.6
6.3
3.8
9.3
-2.9
2.6
-4.6
3.8
8.5
1.4
16
-------
site id(1)
LYK123
LYK123
LYK123
LYK123
LYK123
MCK131
MCK131
MCK231
MKG113
OXF122
OXF122
PAR107
PAR107
PAR107
PAR107
PIN414
PRK134
PSU106
PSU106
SEK430
SND152
SND152
SPD111
STK138
YEL408
PAR107
PSU106
SND152
VIN140
WSP144
MCK231
STK138
STK138
UVL124
PAR107
WSP144
PSU106
SND152
VIN140
WSP144
SEK402
SUM156
VIN140
WSP144
ROM406
VOY413
VIN140
WSP144
UVL124
VIN140
YOS404
N/S distance (km)
203
22
736
242
67
0
-509
-509
-243
49
-87
-181
533
39
-136
6
1676
220
45
0
-1323
-494
-669
-793
394
760
E/W distance (km)
-267
-405
238
359
-650
0
396
396
257
-405
-788
-138
505
626
-383
-191
-449
764
-245
0
-2079
121
-888
-37
-201
744
so2 vd RPD
-1.6
-6.5
-4.9
-8.5
2.7
5.5
7.9
2.4
-2.7
-8.1
-3.8
-5.0
-3.3
-7.0
4.3
-2.3
0.4
-2.0
9.2
-5.4
6.4
-3.7
7.5
1.9
8.4
5.1
hno3 vd RPD
7.5
9.1
3.5
6.6
7.6
2.5
0.8
-1.8
-2.8
-0.9
-0.7
1.6
-4.0
-0.9
0.1
-1.1
3.3
-2.5
-1.5
-0.8
2.6
3.1
4.1
-7.4
9.4
1.0
particulate vd RPD
7.7
9.6
-1.8
5.5
1.9
1.3
2.0
0.7
-1.2
6.2
0.4
1.9
-9.4
-2.1
-5.8
7.5
5.5
-4.1
-7.7
-0.8
7.9
7.3
3.6
9.5
-1.2
6.9
This study also compares the 10-year mean deposition velocities for each site with the annual
mean deposition velocities used to create the 10-year value. Mean Absolute Percent Differences
(MAPD) were calculated to show a summary of the differences, which in a few cases are
between 10 and 15 percent. The results are shown in Table 7. MAPD greater than 10 percent
are color coded with orange shading.
17
-------
Table 7. Mean Absolute Percent Differences between 10-year Mean Deposition Velocities and
Annual Mean Deposition Velocities
site id
ABT147
ACA416
ALC188
ALH157
ANA115
ARE 128
ASH135
ASH235
BBE401
BEL116
BFT142
BVL130
BWR139
CAD150
CAN407
CAT175
CDR119
CDZ171
CHA467
CHE185
CKT136
CND125
CNT169
CON186
COW137
CTH110
CVL151
DCP114
DEN417
DEV412
EGB181
ESP127
EVE419
GAS153
GLR468
GRB41 1
GRC474
GRS420
GTH161
HOW132
HOX148
HVT424
HWF187
IRL141
JOT403
KEF112
KNZ184
LAV410
LRL117
LYE145
LYK123
MAC426
MCK131
MCK231
MEV405
MKG113
MOR409
NCS415
OLY421
OXF122
PAL190
total_years
10
9
4
10
10
10
10
4
10
10
10
10
10
10
10
10
10
10
10
6
10
10
10
5
10
10
10
10
9
10
10
10
9
10
10
10
10
10
10
10
7
6
6
7
10
10
6
10
10
10
10
6
10
10
10
10
10
10
8
10
1
valid_years
5
8
3
6
5
8
8
2
8
8
9
8
7
6
8
2
7
1
8
1
6
7
5
4
10
7
5
7
7
8
8
8
6
5
6
4
9
9
5
5
6
2
4
4
8
8
5
8
5
2
9
5
7
7
8
10
6
5
3
10
0
SO2 MAPD
4.8
9.8
6.6
3.8
4.3
7.6
9.2
8.8
2.1
6.3
6.3
8.0
7.6
2.6
1.8
2.4
4.5
0.0
6.9
0.0
3.9
8.9
3.6
4.1
7.1
5.6
5.0
5.6
13.6
0.3
6.7
5.5
8.2
5.1
5.6
2.0
8.7
4.1
4.8
7.1
11.3
1.2
5.6
2.0
8.5
7.2
6.3
4.7
5.3
2.6
6.8
8.2
8.3
4.9
5.7
5.9
1.4
5.3
0.8
7.9

HNO3 MAPD
2.4
9.9
4.4
3.5
3.1
5.4
2.3
1.2
2.8
4.0
2.2
12.3
2.2
5.1
2.5
12.7
4.9
0.0
1.8
0.0
1.5
3.4
3.3
0.8
2.4
3.0
4.7
2.1
3.8
1.3
2.8
2.9
8.4
2.4
8.1
2.5
10.3
4.7
2.7
4.5
2.3
1.3
2.5
4.3
4.6
4.4
3.0
2.3
3.9
1.2
2.5
0.5
3.4
2.4
4.2
2.6
4.7
4.0
5.2
2.5

PART MAPD
3.1
13.7
5.2
5.0
3.7
7.5
2.3
1.9
4.3
4.9
2.2
11.0
2.6
6.1
3.3
12.5
6.5
0.0
3.1
0.0
3.3
5.3
3.4
3.0
3.1
4.6
4.4
2.0
3.9
2.0
2.8
2.8
6.8
3.4
8.6
3.2
11.9
6.8
2.6
5.8
2.4
1.2
1.4
2.2
5.7
5.9
3.9
3.7
6.6
0.6
3.4
3.2
2.8
3.7
4.2
6.8
7.5
4.0
6.6
3.7

18
-------
site id
PAR107
PED108
PET427
PIN414
PND165
PNF126
POF425
PRK134
PSU106
QAK172
ROM206
ROM406
SAL 133
SAN 189
SEK402
SEK430
SHN418
SND152
SPD111
STK138
SUM156
THR422
UVL124
VII423
VIN140
VOY413
VPI120
WEL149
WNC429
WSP144
WST109
YEL408
YOS404
total_years
10
10
6
10
10
10
3
10
10
10
7
10
10
2
7
3
10
10
10
10
10
9
10
7
10
10
10
1
4
10
10
10
10
valid_years
7
8
3
8
8
5
0
6
6
7
3
8
6
1
2
1
7
7
8
1
6
7
7
3
9
6
7
0
1
6
3
7
4
SO2 MAPD
6.4
6.8
2.6
6.5
2.7
2.2

12.6
9.3
5.7
4.7
5.2
5.1
0.0
1.5
0.0
6.4
7.1
6.5
0.0
5.5
5.4
12.6
1.8
3.8
10.7
7.6

0.0
5.9
5.4
4.9
5.3
HN03 MAPD
2.2
10.7
1.8
2.9
4.5
3.6

9.4
2.4
5.6
1.8
1.9
1.6
0.0
3.1
0.0
2.2
2.0
2.8
0.0
4.6
2.9
6.7
2.3
2.5
1.9
5.6

0.0
2.7
5.8
2.1
1.7
PART MAPD
4.1
14.3
2.2
2.9
5.6
5.4

9.9
3.0
6.1
3.3
2.7
2.0
0.0
4.2
0.0
3.8
4.2
5.5
0.0
6.0
3.2
9.5
3.0
3.3
3.6
7.2

0.0
5.0
7.6
4.7
3.6
Task B:
Using the estimates of uncertainties in the NOs" and HNOs concentrations detailed in the
METHODS section, annual mean fluxes for the Beltsville, MD site (BEL 116) were analyzed to
provide a range of possible values. Total measured nitrogen deposition (NOs'+HNOs+NtLt*)
values were also modified. Because flux estimates are calculated by multiplying the modeled
deposition velocity by the atmospheric concentration, any multiplication of the concentration
essentially applies the same correction to the flux. Table 8 shows the results of the correction.
19
-------
Table 8. Components of the Dry Deposition of Total Measured Nitrogen including Corrections
for Estimates in Concentration Uncertainties
$teJd
BEL116
BEL116
BEL116
BEL116
BEL116
BEL116
BEL116
BEL116
BEL116
BEL116
year,..'
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
,,tto3 fltwc
0.26
0.39
0.49
0.30
0.23
0.28
0.36
0.42
0.27
0.17
•• - xswifitotf- „•:, ••
sio£yjt&(&.5)~,-
0.07
0.11
0.14
0.09
0.07
0.08
0.10
0.12
0.08
0.05
hno^fltflt
9.21
10.53
6.08
8.26
9.12
7.80
8.80
7.86
9.33
8.56
,, '•' CCPfiHJtfld, •- ,
Jwirt3Lflwff145)
10.60
12.11
6.99
9.49
10.48
8.97
10.12
9.03
10.73
9.84
liWJIux
0.66
0.74
0.50
0.65
0.63
0.55
0.67
0.52
0.64
0.59
As Table 9 shows, differences from the corrections are relatively small (approximately 10
percent or less) reflecting the fact that the HNOs component dominates the total measured
nitrogen dry deposition, and it was only modified by increasing the flux by 15 percent. Dry
deposition of particulate NCV has a relatively small contribution to the total measured nitrogen
flux and thus the large decrease in NCV concentrations does not have a significant effect on total
deposition.
Table 9. Changes to Dry Deposition of Total Measured Nitrogen from Corrections to
HNO3 Fluxes
and
siteUch
BEL116
BEL116
BEL116
BEL116
BEL116
BEL116
BEL116
BEL116
BEL116
BEL116
y**K
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
" Total Measured N
2.62
3.00
1.85
2.41
2.57
2.23
2.56
2.25
2.63
2.40
NOQcpraefion
'' :':' .only
2.58
2.94
1.77
2.36
2.53
2.18
2.50
2.18
2.59
2.38
HNO3 correction
only
2.93
3.35
2.05
2.68
2.87
2.49
2.85
2.51
2.94
2.69
both
corrections '
2.89
3.29
1.97
2.64
2.84
2.44
2.79
2.44
2.90
2.66
Task C:
The comparisons of fluxes calculated using 10-year mean deposition velocities and fluxes
calculated with annual mean deposition velocities with historical annual flux values were
evaluated by constructing histograms of percent differences. Comparisons were made using
10-year mean and annual mean values when either three or four quarters are valid (one quarter of
the year may not be represented in the annual mean) and a second set of comparisons were made
using 10-year mean and annual mean values where all four quarters are valid (the entire year is
covered). Figures 6 through 15 present histograms of percent differences versus number of site-
years. There are two histograms for each of five pollutants - SCV HNCV SC>42~, NCV, and
20
-------
NELt+. The first histogram shows results from the comparisons where either three or four
quarters are valid. The second histogram for each pollutant shows results from the comparisons
when all four quarters are valid. Each figure provides two sets of bars. One is based on fluxes
estimated from a 10-year mean (maroon) deposition velocity and the second is based on annual
mean (yellow) deposition velocities.

All ten figures suggest the distributions of differences are normal. The results for 862, HNOs,
NC>3, and NH4+ show a positive bias in that the calculated fluxes are higher than the MLM
results. The exception is for SO42" which shows a slight negative bias. Use of the 10-year mean
allowed for the recovery of 131 site-years (37 percent). Many of the missing site-years are the
start-up or shut-down years for a specific site. Use of the annual mean allowed for the recover y
of only 8 site-years (2 percent) but was also useful in examining how comparable flux estimates
calculated with annual average concentration and deposition velocities are with those aggregated
using the accepted protocol.

The results for Task C show the comparison of SC>2 fluxes as presented in Figure 6 (3 or
4 quarters valid). The simulations based on individual annual Vd values (yellow bars) better
match the MLM results. About 96 percent of the calculated fluxes are within 20 percent of the
MLM deposition rates; and about 53 percent of the calculated values are within 4 percent of the
MLM results. The results based on the 10-year mean deposition velocities are shown by the
maroon bars. Approximately 92 percent of the calculated fluxes are within 20 percent of the
MLM values and about 31 percent are within 4 percent.
21
-------
Figure 6. Histograms of Percent Difference versus Number of Site-Years for
(3 or 4 Quarters Valid)
SO2 - 3 or 4 Quarters Valid

• 10yr mean Vd used
D annual mean Vd used
200
w
•5
100 -
E
50
^l
i
OOOOCDCMOOfOtCOCM
O (0 Tf C\l -^ r; • T-

Percent Difference
S §
Figure 7 shows results for means calculated using years where all four quarters are valid. As
would be expected, results are slightly better as compared with the results using years with either
three or four valid quarters. Approximately 99 percent of the calculated fluxes are within
20 percent of the MLM deposition rates, and about 60 percent of the calculated values are within
4 percent of the MLM results. The results based on the 10-year mean deposition velocities are
shown by the maroon bars. Approximately 95 percent of the calculated fluxes are within
20 percent of the MLM values and about 36 percent are within 4 percent
Figure 7. Histograms of Percent Difference versus Number of Site-Years for
(All Quarters Valid)
SO2 - All Quarters Valid
200 -i

180

160

I 120

J2 100
o
E 80 -
E
Z 60 -|

0
• 10yr mean Vd used
D annual mean Vd used
l
l
§
oo
CD O
Percent Difference
22
-------
The results for fluxes of HNOs, SO42", NOs", and NH4+ are provided in Figures 8 through 15.
The data again show the flux estimates based on individual annual Vd values better match the
MLM results. The fluxes based on 10-year mean Vd values for HNOs, SC>42~, and NH4+ show
an improvement over SC>2. In fact, more than 95 percent of the flux values based on 10-year
mean Vd values are within 20 percent of the MLM annual values. The results for NCV
are generally poor. Particulate nitrate values are generally low and are more uncertain that the
other parameters. For each parameter,
Figure 8. Histograms of Percent Difference versus Number of Site-Years for
(3 or 4 Quarters Valid)
HNO3 - 3 or 4 Quarters Valid
250

200
CD
0
i 150
5
•5
ji 100 -
50 -
n -
• 1 0yr mean Vd used
D annual mean Vd used

III I

JU
O CD tM
CN T- T-
§ 8
Percent Difference
Figure 9. Histograms of Percent Difference versus Number of Site-Years for HNC>3
(All Quarters Valid)
HNO3 - All Quarters Valid
200
tftfi -
I OU
160 -
i 120 -
|ioo-
5 80 -
z 60
40
20-
n -
• luyi mean vu Ubcu
D annual mean Vd used

, , , B~l , B-i , B~i , 1

,11,1-1. •-.,•_,• ,
OOOO
CDCNCOTj-OTtCOCNCDOOOO

Percent Difference
23
-------
Figure 10. Histograms of Percent Difference versus Number of Site-Years for SC>42"

(3 or 4 Quarters Valid)

SO4 - 3 or 4 Quarters Valid

200
O)
Il50
£
w
•5
|j 100
3
50
n
• 1 0yr mean Vd used
D annual mean Vd used

tllll

, In , In . •-. , l~) , .-
OOOOCOCMCO^O^COCMCOOOOO
o o rsi - T- • T- T- CNI Tt co o
Percent Difference
Figure 11. Histograms of Percent Difference versus Number of Site- Years for SC>42~

(All Quarters Valid)
SO4 - All Quarters Valid
200

180
• 10yr mean Vd used

D annual mean Vd used
OTCOCNCOOOOO
T- t- CM •* CO O
Percent Difference
24
-------
Figure 12. Histograms of Percent Difference versus Number of Site-Years for NO3"

(3 or 4 Quarters Valid)
NO3 - 3 or 4 Quarters Valid
250
200
150
100
50
• 10yr mean Vd used

D annual mean Vd used
ial
Percent Difference

Figure 13. Histograms of Percent Difference versus Number of Site-Years for NC>3~

(All Quarters Valid)

NO3 - All Quarters Valid

200
HOyrmean Vd used
-Q

3
180

160

140

120

100

60 -

20
D annual mean Vd used
§OOOCOCMCO'±OT|-eOCMCOOOOO

-------
Figure 14. Histograms of Percent Difference versus Number of Site-Years
(3 or 4 Quarters Valid)
NH4 - 3 or 4 Quarters Valid
ZSU

200
f 150-
55
t5
| 100
E
z
50
0
• 10yr mean Vd used
D annual mean Vd used

\] 1
1 11.1
OOOOCDCMCOTj-o-3-COCMCQOOOO
0 <0 •* CM ^ T- i T- ^ CM T <0 0
Percent Difference
Figure 15. Histograms of Percent Difference versus Number of Site-Years for
(All Quarters Valid)
NH4-All Quarters Valid
200
180
160 -
• 10yr mean Vd used
D annual mean Vd used
Percent Difference
26
-------
Table 10 summarizes the results from Task C. Focusing on the columns for percent of site-years
within +- 20 percent of the historical flux estimate, results are encouraging. For the analysis
done where all four quarters are valid, well over 90 percent of all site-years have a percent
difference within the +- 20 percent window except for NCV, which is slightly lower.

Table 10. Summary of 10-Year Mean and Annual Mean Calculated Flux Comparisons

parameter
so2 10yr flux
pet diff
so2 ann flux
pet diff
hno3 10yr flux
pet diff
hno3_ann_flux
pet diff
so4 10yr flux
pet diff
so4 ann flux
pet diff
no3 10yr flux
pet diff
no3 ann flux
pet diff
nh4 10yr flux
pet diff
nh4 ann_flux
pet diff
•Mjjepeert
;3ior.£,.\.
quarters
31
53
43
60
31
42
24
24
34
55
t
-------
Task D:
Table 11 shows the difference between the current and original parameterizations for the
WATER and ROCK plant types.

Table 11. ROCK/WATER Parameterizations

plwrtjwje ,-,
min stom
light res coeff
opt temp
max no stress temp
min no stress temp
profile index
canopy height
-•; . ,CWfnt;- .
'; v'R|iw^« v
500
20
40
50
20
1
0.1
•;-llfee^
1000
20
30
50
10
1
1
Original
'•• /Pfcra««le![iz9tiQ«; . •
-niwfcreRr.'
1
1
-99
-99
-99
1
0.1
RQGK*
1
1
-99
-99
-99
1
0.1
Comparisons of the weekly deposition velocity estimates are shown in Table 12. The percent
difference is calculated between the actual weekly value and the value calculated using the older,
original parameterization for water and rock. As shown by the table of Mean Absolute Percent
Difference values, there is no difference between the two model runs. The change in
parameterization of water and rock had no effect on the produced deposition velocities.

Table 12. Comparison between Current and Original Parameterizations of WATER and ROCK.
sftejd .
IRL141 (WATER)
CHA467 (ROCK)
Task E:
Model runs were done replacing missing vector wind speed and sigma theta (standard deviation
of the wind direction) data with either historical weekly mean values or "near site" values. For
this part of the study, the "near site" pair of LYK123 and DCP114 was selected with LYK123
being the primary site. Model runs were done for 2007 and then compared with the actual
deposition velocity estimates. For the actual deposition velocity estimates for 2007, LYK123
had a near perfect level of completeness with all 52 weeks having a percent completeness of
greater than 90 percent. This makes this site the perfect candidate for this study.

Table 13 shows a summary of the differences in weekly deposition velocity estimates. Mean
absolute percent differences (MAPD) were calculated for each site from the 52 weekly values.
In general, results were excellent for the comparisons using the historical weekly means. MAPD
values of greater than 5 percent are color coded orange. Only two of the "sites" have MAPD
greater than 5; both involve particulate deposition velocity estimates and the substitution of
sigma theta values for either 10 weeks or all weeks from the first quarter. Results from the runs
that replaced missing data by using hourly values from DCP114 are not as good. Twelve "sites"
have MAPD greater than 5 percent with three sites having MAPD greater than 10 percent.
28
-------
Table 13. Mean Absolute Percent Differences between Weekly Estimates (Test Site vs. Actual)
site id
HISOTG
HISOTS
HISOFG
HISOFS
HISOSG
HISOSS
HISOEG
HISOES
HISONG
HISONS
HISOAG
HISOAS
HISTTG
HISTTS
HISTFG
HISTFS
HISTSG
HISTSS
HISTEG
HISTES
HISTNG
HISTNS
HISTAG
HISTAS
LYKOTG
LYKOTS
LYKOFG
LYKOFS
LYKOSG
LYKOSS
LYKOEG
LYKOES
LYKONG
LYKONS
LYKOAG
LYKOAS
LYKTTG
LYKTTS
LYKTFG
LYKTFS
LYKTSG
LYKTSS
LYKTEG
LYKTES
LYKTNG
LYKTNS
LYKTAG
LYKTAS
SO2 VD MAPD
0.2
0.8
0.6
0.8
0.9
1.5
1.0
2.3
1.4
2.4
1.8
3.3
0.3
0.8
0.5
1.0
0.5
2.2
0.8
2.7
1.0
3.1
1.2
4.3
0.2
0.3
0.6
1.2
1.2
1.5
1.3
2.5
1.7
3.3
2.3
4.1
0.4
0.6
1.1
1.7
1.7
2.6
2.0
2.9
2.5
3.6
3.2
5.0
HNO3 VD MAPD
0.5
0.8
1.7
1.2
2.2
2.6
2.8
3.4
3.7
3.0
4.4
4.9
1.0
0.9
1.2
1.2
1.1
2.5
2.3
2.8
2.7
3.3
3.3
4.7
0.6
0.3
2.3
1.7
3.7
2.1
4.2
2.9
5.5
4.0
7.1
5.1
2.1
1.5
4.3
3.0
5.2
3.5
7.5
4.7
9.5
6.0
12.0
7.9
PARTICULATE VD MAPD
0.6
0.3
2.7
1.1
3.3
2.6
4.2
2.9
5.2
2.1
6.4
3.9
1.0
0.8
1.1
1.3
1.0
1.3
2.6
1.7
2.9
2.5
3.5
2.6
0.9
0.3
4.5
2.0
6.8
2.5
7.7
3.3
10.0
4.5
12.9
5.7
2.2
1.8
4.0
3.3
4.9
3.7
7.6
5.5
9.5
6.8
12.1
8.7
KEY
Q1, 2 Weeks, Sigma theta
Q1 , 2 Weeks, windspeed
Q1, 4 Weeks, sigma theta
Q1, 4 Weeks, windspeed
Q1, 6 Weeks, sigma theta
Q 1 , 6 Weeks , wi nd speed ,
Q1, 8 Weeks, sigma theta
Q1, 8 Weeks, windspeed
Q1, 10 Weeks, sigma theta
Q1, 10 Weeks, windspeed
Q1 , All Weeks, sigma theta
Q1 , All Weeks, windspeed
Q3, 2 Weeks, sigma theta
Q3, 2 Weeks, windspeed
Q3, 4 Weeks, sigma theta
Q3, 4 Weeks, windspeed
Q3, 6 Weeks, sigma theta
Q3, 6 Weeks, windspeed
Q3, 8 Weeks, sigma theta
Q3, 8 Weeks, windspeed
Q3, 10 Weeks, sigma theta
Q3, 10 Weeks, windspeed
Q3, All Weeks, sigma theta
Q3, All Weeks, windspeed
Q1, 2 Weeks, sigma theta
Q1 , 2 Weeks, windspeed
Q1 , 4 Weeks, sigma theta
Q1, 4 Weeks, windspeed
Q1 , 6 Weeks, sigmajheta
Q1 , 6 Weeks, windspeed
Q1 , 8 Weeks, sigma theta
Q1 , 8 Weeks, windspeed
Q1, 10 Weeks, sigma theta
Q1, 10 Weeks, windspeed
Q1, All Weeks, sigma theta
Q1, All Weeks, windspeed
Q3, 2 Weeks, sigma theta
Q3, 2 Weeks, windspeed
Q3, 4 Weeks, sigma theta
Q3, 4 Weeks, windspeed
Q3, 6 Weeks, sigma theta
Q3, 6 Weeks, windspeed
Q3, 8 Weeks, sigma theta
Q3, 8 Weeks, windspeed
Q3, 1 0 Weeks, sigma theta
Q3, 10 Weeks, windspeed
Q3, All Weeks, sigma theta
Q3, All Weeks, windspeed

Table 14 shows the percent difference between annual mean deposition velocities from the actual
run for LYK123 and the 48 test runs. Percent differences less than -5 percent or greater than
5 percent are color coded orange As with the weekly comparison, results from the runs using
historical weekly averages are excellent. Only one site has a percent difference of greater than
5 percent.
29
-------
As with the weekly comparison, it involves the participate deposition velocity estimate and the
substitution of sigma theta values for all weeks from the first quarter. Twelve "sites" where the
missing data are replaced with data from DCP114 have percent differences of less than -
5 percent or greater than 5 percent. For the annual comparisons, one "site", LYKTAS, has
percent differences for SO2 Vd, HNO3 Vd, and particulate Vd are all less than -5 percent.

Table 14. Percent Differences between Annual Estimates (Test Site vs. Actual)
site id
HISOTG
HISOTS
HISOFG
HISOFS
HISOSG
HISOSS
HISOEG
HISOES
HISONG
HISONS
HISOAG
HISOAS
HISTTG
HISTTS
HISTFG
HISTFS
HISTSG
HISTSS
HISTEG
HISTES
HISTNG
HISTNS
HISTAG
HISTAS
LYKOTG
LYKOTS
LYKOFG
LYKOFS
LYKOSG
LYKOSS
LYKOEG
LYKOES
LYKONG
LYKONS
LYKOAG
LYKOAS
LYKTTG
LYKTTS
LYKTFG
LYKTFS
LYKTSG
LYKTSS
LYKTEG
LYKTES
LYKTNG
LYKTNS
LYKTAG
LYKTAS
SO2 VD PCT
0.1
0.7
0.4
0.7
0.8
1.1
0.9
2.0
1.4
2.4
1.8
3.1
0.0
0.5
0.3
0.8
0.1
2.3
0.5
2.7
0.7
3.1
0.9
4.4
0.1
0.0
0.5
-1.3
1.1
-1.2
1.2
-2.0
1.7
-3.1
2.3
-3.8
0.1
-0.5
1.1
-1.8
1.8
-3.0
2.1
-3.3
2.7
-4.0
3.5
-5.5
HNO3 VD PCT
0.3
0.6
1.5
0.6
1.9
1.3
2.4
2.0
3.2
2.1
3.9
3.2
0.7
0.7
1.2
0.2
0.1
2.2
1.5
2.6
2.0
3.1
2.2
4.5
0.4
-0.1
2.0
-1.6
3.2
-1.6
3.7
-2.5
4.8
-3.5
6.2
-4.3
2.4
-1.7
4.6
-3.3
5.1
-3.6
7.7
-5.0
9.9
-6.4
12.6
-8.4
PARTICULATE VD PCT
0.3
0.1
2.0
-0.2
2.9
0.4
3.6
0.7
4.3
-0.1
5.4
0.9
1.0
-1.1
1.3
-1.7
0.1
0.0
1.9
-1.2
2.2
-1.6
2.4
-1.5
0.5
-0.2
3.5
-1.6
6.3
-2.1
6.8
-3.1
8.5
-4.0
11.6
-5.3
3.0
-2.5
4.7
-3.9
4.9
-3.8
8.1
-6.0
10.2
-7.6
12.8
-9.6
KEY
Q1 , 2 Weeks, sigma theta
Q1 , 2 Weeks, windspeed
Q1, 4 Weeks, sigma theta
Q1 , 4 Weeks, windspeed
Q1, 6 Weeks, sigmajheta
Q1 , 6 Weeks, windspeed
Q1, 8 Weeks, sigma theta
Q1 , 8 Weeks, windspeed
Q1 , 10 Weeks, sigma theta
Q1 , 10 Weeks, windspeed
Q1, All Weeks, sigma theta
Q1 , All Weeks, windspeed
Q3, 2 Weeks, sigmajheta
Q3, 2 Weeks, windspeed
Q3, 4 Weeks, sigma_theta
Q3, 4 Weeks, windspeed
Q3, 6 Weeks, sigmajheta
Q3, 6 Weeks, windspeed
Q3, 8 Weeks, sigma theta
Q3, 8 Weeks, windspeed
Q3, 10 Weeks, sigma theta
Q3, 10 Weeks, windspeed
Q3, All Weeks, sigma theta
Q3, All Weeks, windspeed
Q1, 2 Weeks, sigma theta
Q1, 2 Weeks, windspeed
Q1, 4 Weeks, sigma theta
Q1 , 4 Weeks, windspeed
Q1, 6 Weeks, sigma theta
Q1 , 6 Weeks, windspeed
Q1, 8 Weeks, sigma theta
Q1 , 8 Weeks, windspeed
Q1, 10 Weeks, sigma theta
Q1, 10 Weeks, windspeed
Q1, All Weeks, sigma theta
Q1 , All Weeks, windspeed
Q3, 2 Weeks, sigma theta
Q3, 2 Weeks, windspeed
Q3, 4 Weeks, sigma theta
Q3, 4 Weeks, windspeed
Q3, 6 Weeks, sigma theta
Q3, 6 Weeks, windspeed
Q3, 8 Weeks, sigma theta
Q3, 8 Weeks, windspeed
Q3, 10 Weeks, sigma theta
Q3, 10 Weeks, windspeed
Q3, All Weeks, sigma theta
Q3, All Weeks, windspeed
30
-------
CONCLUSIONS

Task A:
The analysis of the comparison between 10-year mean deposition velocities shows that the
practice of determining "near sites" based on geographic proximity and similar terrain and land
use is not supported by a comparison of the data produced by the model. Of the 89 site pairs
identified by looking for matches where the S02 Vd, HNOs Vd, and particulate Vd are all within
±10 percent, only five are currently used as "near sites" and most of the site pairs identified were
not near one another. From this analysis, it is recommended that the use of "near sites" for
missing data replacement be terminated.

Task B:
Previous studies have shown that CASTNET NO{ and HNC>3 measurements have an associated
uncertainty due to sampler design. Differences from the corrections enacted to examine the
effects of these uncertainties are relatively small because the HNOs component dominates the
total measured nitrogen dry deposition. Dry deposition of particulate NCV has a small
contribution to the total measured nitrogen flux and thus the large decrease to NOs"
concentrations does not have a significant effect on the total measured nitrogen estimate. Similar
to the conclusions of Lavery et al.3 regarding total measured nitrogen concentrations, the dry
deposition estimates of total measured nitrogen concentrations are'reasonable despite the
problems with NCV and HNCh collection.

Task C:
Evaluation of the results in this Task shows that flux estimates calculated using 10-year mean
deposition velocities and annual mean concentrations compare reasonably well. Most
comparisons had a percent completeness between -20 and 20 percent. Results obtained using
10-year mean deposition velocities calculated using only years with four valid quarters were
better than those calculated using all valid site-years (either three or four valid quarters). These
results are encouraging and show that reasonable approximations-of annual deposition estimates
might be obtained using a summary deposition velocity such as a 10-year mean.

Task D:
Changes to the parameterization for the WATER and ROCK plant types had no effect on the
deposition velocity estimates.

Task E.
The most promising results of this study come from the replacement of missing meteorological
parameters (specifically vector wind speed and sigma theta) with historical weekly averages of
these parameters using data from the same site. These model runs were much better than the
results obtained by substituting for missing data using "near site" meteorological data.
31
-------
The chief recommendation from this study is to proceed with the development of a missing or
invalid data replacement scheme based on:
1) Use of a summary, 10-year mean deposition velocity, and
2) Use of historical weekly averages for all meteorological inputs used by the MLM.

Results from Tasks C and E show that both protocols have enough promise to warrant further
study and investigation. The potential benefit is enormous in that all site-years missing due to a
lack of available valid meteorological data could be recovered. Details of the replacement
protocol will require further development including how many years are required, what is the
backup to weekly averages, and what to do for missing data at new sites without a historical
record. Similar issues were successfully dealt with when CASTNET converted to calculation
atmospheric concentrations using local conditions (as opposed to standard conditions). The
temperature replacement protocol could be viewed as a model of what has been done in the past
to deal with missing or invalid meteorological data.

This study only examined a single site for one year. Further work in support of the development
and testing of the final replacement protocol should perform a similar analysis on all sites for
multiple years. Also, it would be beneficial to reconstruct Figures 3-3 thru 3-10 from the
CASTNET 2007 Annual Report1' using the results of Tasks C and E and review the differences.
If results of the time series analyses are approximately the same, the finding would lend support
and validation to the adoption of the replacement scheme.

DISCLAIMER

This report was prepared under EPA contract EP-D-05-096 and may be freely distributed and
used for non-commercial, scientific and educational purposes. The views and opinions of
authors expressed herein do not state or reflect those of the United States Government, and shall
not be used for advertising or product endorsement purposes.

REFERENCES

1. Cooter, E.J.; Schwede, D.B. J. Geophys. Res., 2000, 705, 6695-6707.

2. Rogers, C.M; Lavery, T.F.; Barnard, W.R.; Howell, H.K.; Sartain, R.T. 2003, Estimating
Deposition Velocities for the Clean Air Status and Trends Network (CASTNET) using
the Multi-layer Model (MLM): Sensitivity to Model Inputs. Proceedings of the AWMA
Annual Meeting, San Diego, California, June 22-26.

3. Lavery, T.F.; Rogers, C.M.; Baumgardner, R.E., Jr.; Mishoe, K.P. J. Air & Waste
Manage. Assoc., In Press, Intercomparison of CASTNET NOs" and HNOs Measurements
with Data from Other Monitoring Programs.
32
-------
4. CASTNET Home Page, http://www.epa.gov/castnet/ (accessed September 2008).

5. Meyers, T. P.; Finkelstein P. L.; Clarke J.; Ellestad T. G.; Sims P. J. Geophys. Res. 1998,
705,22,645-22,661.

6. Finkelstein, P.L., Ellestad, T.G., Clarke, J.F., Meyers, T.P., Schwede, D.B., Hebert, E.G.,
and Neal, J.A. J. Geophys. Res. 2000,105, 15,365-15,377.

7. Hicks, B. B.; Baldocchi, D. D.; Hosker, R. P., Jr.; Hutchison, B. A.; McMillen, R. T.;
Satterfield, L. C. On the Use of Monitored Air Concentrations to Infer Dry Deposition;
NOAA Technical Memorandum ERL ARL-1411985; National Oceanic and Atmospheric
Administration, Air Research Laboratory: Silver Springs, MD, 1985.

8. Meyers, T. P.; Yuen, T. S. J. Geophysic Res. 1987, 92, 6705-6712.

9. Schwede, D.B. 2006. A Comparison of the Deposition Velocity Estimates from the
CASTNET and CAPMoNNetworks. (Working paper). Research Triangle Park, NC.

10. MACTEC Engineering and Consulting, Inc. (MACTEC). 2008. Clean Air Status and
Trends Network (CASTNET) Quality Assurance Project Plan (QAPP), Revision 4.1.
Prepared for U.S. Environmental Protection Agency (EPA), Washington, D.C., Contract
No. 68-D-98-112. Gainesville, FL.

11. MACTEC Engineering and Consulting, Inc. (MACTEC). 2008. Clean Air Status and
Trends Network (CASTNET) 2008 Annual Report. Prepared for U.S. Environmental
Protection Agency (EPA), Washington, D.C., Contract No. 68-D-98-112. Gainesville,
FL.
33
-------