Content uploaded by Lowell L Ashbaugh
Author content
All content in this area was uploaded by Lowell L Ashbaugh on Jan 31, 2015
Content may be subject to copyright.
Paper No. 32 “Efficiency of IMPROVE Network Denuders for Removing Nitric Acid”
Regional and Global Perspectives on Haze: Causes, Consequences and Controversies-
Visibility Specialty Conference, Air & Waste Management Association, Asheville, NC
October 25-29, 2004
Efficiency of IMPROVE Network Denuders for
Removing Nitric Acid
Lowell L. Ashbaugh, Charles E. McDade, Warren H. White, Paul Wakabayashi
Crocker Nuclear Laboratory
University of California
One Shields Avenue
Davis, California, 95616 USA
Jeffrey L. Collett, Jr., Xiao-Ying Yu
Atmospheric Science Department
Colorado State University
Fort Collins, CO 80523 USA
Paper #32
ABSTRACT
The IMPROVE (Interagency Monitoring of Protected Visual Environments) network
collects fine particles on nylon filters for analysis of ions, including NO3-. An annular
anodized aluminum denuder coated with Na2CO3 and glycerin (to prevent drying and
improve efficiency) precedes the filters to remove nitric acid. The denuders are changed
once each year during annual maintenance, but otherwise are not serviced. If the efficiency
of the denuder for removing nitric acid decreases during the year, it could allow nitric acid
to be collected on the nylon filter. This would result in a positive artifact, as the nitric acid
would be interpreted as particulate nitrate.
Several changes have been made to the IMPROVE ion module since the first samples were
collected. It’s possible that these changes may have affected collection of nitrate, also. To
examine the efficiency of the denuder in several configurations for removing nitric acid, we
collected daily samples for four consecutive weeks at Brigantine National Wildlife Refuge,
Grand Canyon National Park, and San Gorgonio Wilderness. At San Gorgonio, we
collected samples during two separate four-week periods.
The results of this study show that the denuder in the IMPROVE ion module effectively
and efficiently removes nitric acid.
INTRODUCTION
The IMPROVE monitoring network was established in 1987 to identify the components of
particulate matter that reduce visibility and to track visibility trends over time.1 The
IMPROVE sampler collects PM2.5 (particles with aerodynamic diameter less than 2.5 µm)
on three different substrates, each optimized for the type of analysis performed on it.
Module A collects particles on Teflon filters for mass, light absorption, and elemental
analysis by gravimetry, laser transmission/reflectance, and X-ray fluorescence. Module B
collects particles on nylon filters for ion analysis by ion chromatography, and Module C
collects particles on quartz filters for carbon analysis by thermal optical reflectance.
The nylon substrate on Module B was chosen to retain nitrate ion when ammonium nitrate
dissociates, as it is known to do under conditions of high temperature and low relative
humidity. Nylon filters also trap and retain nitric acid, though, so the sampler was fitted
with an annular anodized aluminum denuder to remove it. Initially, the denuder was coated
with sodium carbonate to ensure high nitric acid collection efficiency.
Several sampling changes have occurred on the ion module since the IMPROVE network
was established. Until summer 1994 the module collected particles on 47mm nylon filters.
From that time until summer 2000, a 25mm nylon filter was used. This filter suffered from
a high pressure drop and frequently clogged, so a 37mm filter has been used since summer
2000. Originally, the denuder was coated with sodium carbonate alone, but glycerin was
added to the coating in summer 1996 to prevent drying. Different filter manufacturers and
filter lots have been used, too, as the filter is a consumable item. Filter lots usually change
about once per year, but the manufacturer changed in fall 1996 and again in 2004.
Typically, nitrate concentrations are higher and more variable during the winter than during
the summer. From 1996 to 2000, many sites (though not all) showed lower winter nitrate
concentrations along with lower variability.
Figure 1 illustrates this phenomenon at Mammoth Cave, and also shows the timing of some
network changes on the ion module. One hypothesis for the phenomenon was that the
denuder became more efficient at removing nitric acid when glycerin was added. If true,
this would mean that the earlier nitrate concentrations were artificially high and could not
be used for trends analysis. But there is no corresponding explanation for the fact that the
variation and general level of concentrations returned to “normal” in 2000-01. It’s
important for tracking progress under the Regional Haze Rule2 to understand whether this
change in particle collection is due to changes in operation of the network or due to real
changes in the atmosphere.
Figure 1. Nitrate concentrations at Mammoth Cave National Park. Network changes
for the ion module are also shown.
Mammoth Cave
-1
0
1
2
3
4
5
6
7
8
3/1/93 3/1/94 3/1/95 3/1/96 3/1/97 3/1/98 3/1/99 3/1/00 3/1/01 3/1/02
NO3, ug/m3
changed IC lab added glycerin to denuder changed filter mfr
moved site 7 miles intro'd version 2 sampler DI water extraction
47 mm filter 25 mm filter 37 mm filter
This study was established to test the efficiency of IMPROVE denuders under a range of
conditions, including different seasons and geographic locations. The sites were chosen to
challenge the sampler with high nitrate and/or nitric acid concentrations.
EXPERIMENTAL METHODS
We collected daily samples for four consecutive weeks at Brigantine National Wildlife
Refuge (Atlantic coast), Grand Canyon National Park (southwest), and San Gorgonio
Wilderness (southern California). At San Gorgonio, we collected samples during two
separate four-week periods. At each site, we operated five IMPROVE ion modules, each
with a different denuder configuration; 1) a freshly coated denuder, 2) a denuder that had
operated for a year at Joshua Tree National Monument (and was exposed to high levels of
nitric acid during that time), 3) an anodized aluminum denuder with no coating, 4) a
Na2CO3 -coated denuder without glycerin, and 5) no denuder. For the module with no
denuder, the bare aluminum surfaces of the inlet tube and rain shield could remove some
nitric acid.
The freshly coated denuder represents an IMPROVE site immediately after annual
maintenance. The used denuder was selected to represent a worst-case exposed denuder at
the end of its annual cycle. The denuder without glycerin was used in the IMPROVE
network prior to summer 1996. Bare aluminum denuders and the bare aluminum surfaces
of sampler inlets have been shown to collect nitric acid3, so these configurations were
included to complete the test.
In addition, two complete suites of Teflon, nylon, and quartz module samplers were run to
collect both PM10 and PM2.5 particles. Full IMPROVE speciation was carried out on each
size range, including particle mass, light absorption, elemental content, major anions and
cations, and carbon fractions. Ion chromatography analysis was carried out at Research
Triangle Institute (RTI) on all nylon filters to measure concentrations of NO3-, SO4=, Cl-,
NH4+, Na+, Mg++, K+, and Ca++. Figure 2 shows the sampling site and the sampler
installation at Brigantine.
IMPROVE sampler filters were changed on-site twice per week and held there until the end
of the sampling period. At that time, all samples were shipped back to UC Davis for further
processing. At UC Davis, the samples were logged in, and then sent to RTI for analysis. All
data processing was conducted using computer spreadsheets developed for that purpose.
Nitric acid and particle nitrate concentrations were measured independently by Colorado
State University using an annular denuder/filter-pack sampler manufactured by University
Research Glassware (URG).4 The sampler consisted of a PM2.5 cyclone inlet, two annular
denuders (with suitable coatings to collect gaseous SO2 and HNO3 in one denuder and NH3
in the other), a two-stage filter-pack housing two nylon filters (Gelman Nylasorb, 37 mm)
in series, and a backup denuder downstream of the filter-pack to collect NH3 volatilized
from particles collected on the first nylon filter. The second nylon filter was included to
trap any volatilized nitric acid that was not recaptured by the first nylon filter. Most of the
URG samples were collected over 24 hr periods, from 08:00 to 08:00 local time.
Figure 2. Photograph of the Brigantine sampling site showing the samplers used in
this study. The trailer housed all the denuder test modules, as well as the PM2.5 and
PM10 speciation modules.
Samples were collected from 8:00 a.m. to 8:00 a.m. each day for approximately four weeks
at each site. Brigantine operated from November 5-30, 2003. Grand Canyon operated from
May 1-30, 2003. San Gorgonio operated during two periods: April 5-26, 2003 and June 30-
July 30, 2003. One field blank was collected each week from each sampling module; the
median field blank measurement for a site was subtracted from each ambient measurement
to correct for artifacts.
The URG denuders were extracted daily using deionized water. The nylon filters were
unloaded in an ammonia-free glove box and stored frozen (with NH3-removing towels)
until later extraction and analysis in the laboratory at Colorado State University. The first
stage nylon filter was extracted using deionized water; the backup nylon filter was
extracted using a basic 1.7 mM sodium bicarbonate / 1.8 mM sodium carbonate solution.
All filters were sonicated during extraction.
Ion analysis at Colorado State University was completed on two Dionex DX-500 ion
chromatographs to measure anion (NO3-, Cl-, and SO42-) and cation (Na+, NH4+, K+, Mg2+,
and Ca2+) concentrations. Detection was by conductivity. Both ion chromatographs were
calibrated daily using a series of standards prepared from analytical grade salts. Replicate
injections and analysis of independent NIST traceable standards were used to establish
measurement precision and accuracy. Uncertainties for major aerosol species
concentrations were found to be of the order of several percent. Additional details on the
analytical procedure can be found in Collett et al. (2004).
RESULTS AND DISCUSSION
Figure 3 shows the nitrate concentrations measured for each of the denuder configurations
tested at each site, along with the nitric acid concentrations measured by Colorado State
University. Note the difference in scale for each figure. The nitric acid concentrations were
comparable for three sites/sample periods, but were significantly higher at San Gorgonio
during July. The particle nitrate concentrations at San Gorgonio in April were an order of
magnitude higher than at Grand Canyon, and were intermediate at Brigantine and San
Gorgonio in July. At all sites, though, there is no indication of a difference in nitrate
collection that depends on denuder configuration. Even in the sampler without a separate
denuder the aluminum walls of the inlet tube efficiently removed the available nitric acid.
Figure 3. Particle nitrate at each site/time period for each denuder configuration. Also
shown is the nitric acid concentration measured by Colorado State University. Note
the different concentration scale for each plot.
San Gorgonio - April
0
2000
4000
6000
8000
10000
12000
14000
16000
4/5/03
4/7/03
4/9/03
4/11/03
4/13/03
4/15/03
4/17/03
4/19/03
4/21/03
4/23/03
4/25/03
ng/m3
None
New
Bare Al
No glyc
Used
HNO3(g)
URG NO3-
Grand Canyon
0
100
200
300
400
500
600
700
800
900
1000
5/1/03
5/3/03
5/5/03
5/7/03
5/9/03
5/11/03
5/13/03
5/15/03
5/17/03
5/19/03
5/21/03
5/23/03
5/25/03
5/27/03
5/29/03
5/31/03
ng/m3
None
New
Bare Al
No glyc
Used
HNO3(g)
URG NO3-
San Gorgonio - July
0
1000
2000
3000
4000
5000
6000
7000
8000
6/30/03
7/2/03
7/4/03
7/6/03
7/8/03
7/10/03
7/12/03
7/14/03
7/16/03
7/18/03
7/20/03
7/22/03
7/24/03
7/26/03
7/28/03
7/30/03
ng/m3
None
New
Bare Al
No glyc
Used
HNO3(g)
URG NO3-
Brigantine
0
500
1000
1500
2000
2500
3000
3500
11/4/03
11/6/03
11/8/03
11/10/03
11/12/03
11/14/03
11/16/03
11/18/03
11/20/03
11/22/03
11/24/03
11/26/03
11/28/03
11/30/03
ng/m3
None
New
Bare Al
No glyc
Used
HNO3(g)
Table 1 shows the mean concentrations of nitrate measured by each denuder configuration
for each site/time period. The five measurements using IMPROVE samplers are very
consistent with each other. They differ slightly from the measurements by CSU using the
URG sampler.
Table 1. Average concentration of nitrate and nitric acid at each test site during the
sampling period.
CSU (µg/m3) UCD NO3- (µg/m3)
HNO3 NO3- No
denuder New
denuder Bare
aluminum No
glycerin
Used
denuder
Brigantine (November) 725 957 969 973 985 957
San Gorgonio (April) 419 3231 3267 3223 3212 3108 3167
Grand Canyon (May) 562 300 245 238 243 237 238
San Gorgonio (July) 4262 1623 1752 1748 1742 1749 1805
The nitrate measurements for each IMPROVE sampler were compared to the CSU URG
sampler using regression analysis. Figure 4 and Table 2 show the slope, intercept, and
correlation coefficients for the regression of each IMPROVE nitrate measurement on the
URG nitrate measurements. A slope greater than one indicates the IMPROVE denuder is
less efficient than the URG denuder. An intercept different from zero suggests there may
be a bias in one measurement or the other, perhaps due to artifact correction. All the
comparisons show high correlations, with the most scatter shown in the Grand Canyon
measurements where correlations were lowest.
Figure 4. Regression of IMPROVE measurements on CSU URG measurements of
nitrate for San Gorgonio and Grand Canyon. At Brigantine the comparison is
between individual IMPROVE measurements and the mean of all IMPROVE
measurements.
San Gorgonio, April 2003
y = 0.948x + 119.045
R2 = 0.996
y = 0.948x + 163.848
R2 = 0.996
y = 0.939x + 137.224
R2 = 0.996
y = 0.912 x + 120.517
R2 = 0.997
y = 0.921 x + 151.005
R2 = 0.996
0
3000
6000
9000
12000
15000
03000 6000 9000 12000 15000
CSU URG Nitrate (µg/m3)
IMPROVE Nitrate (µg/m3)
None
New
Bare
No Glyc
Used
Grand Canyon, May 2003
y = 1.032x - 63.119
R2 = 0.975
y = 0.992x - 44.354
R2 = 0.973
y = 0.978x - 50.338
R2 = 0.959
y = 0.999x - 54.008
R2 = 0.965
y = 1.010x - 56.341
R2 = 0.951
0
100
200
300
400
500
600
0100 200 300 400 500 600
CSU URG Nitrate (µg/m3)
IMPROVE Nitrate (µg/m3)
None
New
Bare
No Glyc
Used
San Gorgonio, July 2003
y = 1.004x + 144.238
R2 = 0.973
y = 1.002x + 153.330
R2 = 0.968
y = 1.032x + 96.493
R2 = 0.968
y = 1.052x + 67.607
R2 = 0.978
y = 1.100x + 48.323
R2 = 0.984
0
1000
2000
3000
4000
5000
6000
01000 2000 3000 4000 5000 6000
CSU URG Nitrate (µg/m3)
IMPROVE Nitrate (µg/m3)
None
New
Bare
No Glyc
Used
Brigantine, November 2003
y = 0.978x + 10.485
R2 = 0.999
y = 0.977x + 22.834
R2 = 0.997
y = 0.998x + 6.429
R2 = 0.999
y = 1.023x - 5.371
R2 = 0.999
y = 1.024x - 34.377
R2 = 0.994
0
700
1400
2100
2800
3500
0700 1400 2100 2800 3500
Mean IMPROVE Nitrate (µg/m3)
IMPROVE Nitrate (µg/m3)
None
New
Bare
No Glyc
Used
There does not appear to be a systematic difference between the denuder options for any of
the test periods, except perhaps for San Gorgonio in July. The anodized aluminum surfaces
of the sampler rain shield and inlet tube (i.e. no denuder) remove nitric acid as efficiently
as the new denuder coated with Na2CO3 and glycerin, even at San Gorgonio in July. The
denuder that had been used for a year prior to testing may have shown some reduced
efficiency in the last sample period at San Gorgonio in July, but was as efficient as the new
denuder until then. It should be noted that the four one-month-long test periods were
equivalent to approximately a one-year operational period for the IMPROVE network.
When the San Gorgonio regressions for July were forced to have a zero intercept (data not
shown), all but the used denuder showed slopes of 1.07. The used denuder showed a slope
of 1.12.
Table 2. Regression coefficients for IMPROVE particulate nitrate measurements
regressed against CSU URG particulate nitrate measurements.
Slope No
denuder New
denuder Anodized,
no coating
No
glycerin Used
denuder
Brigantine NA NA NA NA NA
San Gorgonio (April) 0.948 0.948 0.939 0.912 0.921
Grand Canyon 0.992 1.032 0.978 0.999 1.010
San Gorgonio (July) 1.002 1.004 1.032 1.052 1.100
Intercept
(µg/m3) No
denuder New
denuder Anodized,
no coating
No
glycerin Used
denuder
Brigantine NA NA NA NA NA
San Gorgonio (April) 163.8 119.0 137.2 120.5 151.0
Grand Canyon -44.4 -63.1 -50.3 -54.0 -56.3
San Gorgonio (July) 153.3 144.2 96.5 67.6 48.3
Correlation (r2) No
denuder New
denuder Anodized,
no coating
No
glycerin Used
denuder
Brigantine NA NA NA NA NA
San Gorgonio (April) 0.998 0.998 0.998 0.999 0.998
Grand Canyon 0.986 0.987 0.980 0.983 0.975
San Gorgonio (July) 0.984 0.986 0.984 0.989 0.992
CONCLUSIONS
The denuder used in the IMPROVE network operates efficiently to remove nitric acid
during the one-year period it remains in the field. The sodium carbonate coating is not
necessary for proper operation in most cases, nor is glycerin required to maintain
efficiency. It’s possible that a denuder exposed to two years equivalent of IMPROVE
sampling shows reduced efficiency for removing nitric acid, but it’s not clear that this is the
case.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the field efforts of Taehyoung Lee to collect the
samples used in this study. Funding was provided by the National Park Service through the
IMPROVE Steering Committee under contract number C2350990001.
REFERENCES
1. Eldred, R. A.; Cahill T. A.; Pitchford M. L.; and Malm W. C., Air and Waste
Management Association, 81st Annual Meeting. 1987. Dallas, TX: Air and Waste
Management Association.
2. Guidance for Tracking Progress Under the Regional Haze Rule, EPA-454/B-03-004
(2003), www.epa.gov/ttn/oarpg/t1/memoranda/rh_tpurhr_gd.pdf, accessed September
2004.
3. John, W.; Wall, S. M.; and Ondo, J. L., Atmos. Environ. 1988, 22(8), 1627-1636.
4. Collett, Jr., J. L.; Lee, T., Yu, X-Y.; Ayres, B.; and Kreidenweis, S. M., Air and Waste
Management Association, 97th Annual Meeting, Dallas, TX, Air and Waste
Management Association, June 22-25, 2004.