Georgia Institute of Technology – School of Earth & Atmospheric Sciences

Continuous Gas Measurements During the Fall-Line Air Quality Study 2000

Karsten Baumann

The following is a descriptive summary of continuous gas-phase measurements made during the FAQS 2000 phase I pilot field study at Macon, Sandy Beach Park (6/11 – 6/21), Augusta, Fort Gordon (6/25 – 7/10), and Columbus, Water Works facility (7/13 – 7/23). The sample gas for the measurement of ambient levels of NO, NOy, O3, CO, and SO2 was drawn through PFA tubing with inlets mounted on top of a standard meteorological triangular Al tower, about 9 m above ground. The data were acquired at 1 Hz and reported as 1 min averages if not noted otherwise. Samples of O3, CO, and SO2 were tapped off a manifold inside the trailer at the base of the tower. The manifold was connected to a 11 m long, 9.5 mm ID Teflon-PFA tube through which ambient air was sampled at an average flow rate of 25 slm (liters per min @STP). The average residence time of the sample gas in the line was 1.4 s. Based on earlier tests with addition of CO and SO2 standards to the inlet, chemical losses of these species due to adsorption to the internal surface of the tube wall were not occurring. The quality of the measured data is assessed below in terms of response time, detection limit, precision and accuracy. Values for detection limit and precision are expressed at 95 % confidence level assuming normal distribution of zero and span check signals acquired during each city’s measurement period.

Ozone

O3 was measured using a pressure and temperature compensated commercial UV absorption instrument (model TEI 49-C, Thermo Environmental Instruments, Inc., Franklin, MA), being absolutely calibrated by the known absorption coefficient of O3 at 254 nm. The signal was generated by the difference of frequently alternating measure and reference (zero) cycles, i.e. full transfer of O3 through versus complete removal of O3 from the flow system. The linearity and precision of the analyzer was checked once every 22 hours. Precision check mixing ratios of 0, 40, 80, 120, and 160 ppbv were provided by a primary standard calibrator with active feedback control (model TEI 49C-PS). The calibrator was supplied with O3-free (zero) air from a cartridge of activated carbon that effectively removed O3 from the ambient air. Each precision check resulted in a 5 point linear regression. Assuming normal distribution of the regressions’ intercepts, the O3 analyzer’s detection limits for all 3 sites were 0.5 to 0.6 ppbv; whereas the slopes of the linear regressions yielded ±4 to ±7 % precision. The accuracy is estimated to be the same.

Carbon monoxide

CO was measured by gas filter correlation, nondispersive infrared absorption (model TEI 48C-TL with a hand-selected PbSe detector matched with an optimal preamplifier, and an absorption cell with gold-plated mirrors). The signal output was pressure compensated while the absorption cell temperature was controlled at 42 ±0.3 oC during the entire study. The instrument was modified according to Parrish et al. [1994] by introducing a zero trap of 0.5 % Pd on alumina catalyst bed (type E221 P/D, Degussa Corp.) kept at 180 oC that quantitatively oxidized CO to CO2 and allowed frequent switching between measure and zero modes. Ambient water vapor was previously found to alter the zero level due to the capacity of the zero trap for water absorption. This problem was largely eliminated here, since a small fraction of the sample stream (<1 %) was drawn through the trap (50 cm3 volume) during measure mode. This sufficiently conditioned the zero trap to the variing ambient water vapor levels. Nevertheless, the instrument was switched into zero mode every 11 min for 2 min. This allowed frequent determination of full 1 min zero averages, since the instrument’s response time was 20 s. Careful balance of the two flow schemes prevented any noticeable pressure differences in the absorption cell. NIST traceable calibration gas of 405 ±4 ppmv CO in N2 (Scott-Marrin Inc., Riverside, CA) was introduced into the sample stream by mass flow controlled standard addition and dynamic dilution at the instrument inlet for 2 min approximately every 7 hours. The valve sequence was programmed such that standard addition coincided with a zero about once every day allowing quantification of the zero trap efficiency, which resulted in CO removal > 99 % at all times. The detection limit for a 1 min average based on the 1 Hz data ranged between 94 and 106 ppbv, and between 6 and 9 ppbv for a 1 hour average. The instrument’s precision, determined from the standard addition span checks, ranged between 8 and 14 % at ~600 ppbv. The accuracy was estimated as the RMS error of uncertainties in the calibration tank concentration (2 %), the mass flow controllers (4 % each MFC), the background variation (4 %), and potential inaccuracies from interpolation of the measured ambient CO during span checks (15 %). Thus, the total uncertainty in the CO measurement is estimated at ±17 % for the entire measuring range. The instrument’s linearity within its 5000 ppbv range was checked with a 4 point calibration (zero excluded) at the beginning of the study, and revealed an r2 of 0.9982.

Sulfur dioxide

SO2 was measured by use of a commercial, pulsed UV fluorescence instrument (model TEI 43C-TL) with pressure and temperature compensated signal output. It’s response time was ~45 s and therefore, required longer zeroing and calibration periods compared to the CO instrument: zero for 4 min once every 55 min; calibration - via mass flow controlled standard addition of 30.6 ±0.3 ppmv SO2 in N2 NIST traceable calibration gas (Scott-Marrin Inc.) and dynamic dilution at the instrument inlet - was performed for 4 min once every 330 min. This sequence caused the standard addition to coincide with the zero about every 14 hours. Zero [SO2-free] air was produced by passing ambient air through a HEPA glass fiber in-line filter (Balston) impregnated with a 0.15 molar Na2CO3 solution. At a flow rate of 0.9 slm, the filter removed 100 % of the SO2 in the sample. The instrument exhibited a relatively large sensitivity to ambient temperature variations inside the mobile laboratory, which were monitored via a RTD temperature sensor next to the instrument’s position, and corrected by linear regression with the background signal. Calibrations were performed and evaluated analogous to the CO measurements resulting in a detection limit of 0.19 to 0.27 ppbv, and a precision of ±4 to 9 % at ~60 ppbv. Since the instrument’s measurement principle is known to be sensitive to organic hydrocarbons (HC), the efficiency of the internal HC removal through a semi-permeable wall was enhanced by introducing an activated carbon trap into the flow of the low-[HC]-side of the wall, and thereby further increasing the [HC] gradient across the wall. NO is known to be another interferent, and its level of interference was examined by standard addition of NO calibration gas earlier before the study, resulting in a 2-3 % increase of signal. The reported data were not corrected for this relatively small interference. The accuracy was estimated as the RMS error of uncertainties in the calibration tank concentration (2 %), the mass flow controllers (4 % each MFC), the background variation (12 %), the NO interference (2 %), and potential inaccuracies from interpolation of the measured ambient SO2 during span checks (10 %). Thus, the total uncertainty in the SO2 measurement is estimated at ±17 % for the entire measuring range. The instrument’s linearity within its 100 ppbv range was checked with a 4 point calibration at the beginning of the study, and revealed an r2 of 0.9998.

Nitrogen oxide and sum of total reactive nitrogen oxides

A proto-type Air Quality Design (AQD, Golden, Colorado) NO/NOy analyzer was deployed for the measurement of NO and total reactive nitrogen oxides (NOy) that include NO, NO2, NO3, N2O5, HONO, HNO3, aerosol nitrate, PAN and other organic nitrates. These measurements were based on the principal method of metal-surface induced reduction of the more highly oxidized species to NO [Fahey et al. 1985; 1986; Fehsenfeld et al. 1987; Atlas et al. 1992; Parrish et al. 1993; etc.], and its subsequent chemiluminescence detection (CLD) with excess ozone [Ridley and Howlett 1974; Kley and McFarland 1980; Bollinger 1982; Fehsenfeld et al. 1990]. The metal surface here was a 35 cm long, 0.48 cm ID MoO tube (Rembar Co., Dobbs Ferry, NY), temperature controlled at 330 ±2 oC, and housed inside an inlet box mounted to the met tower ~8 m above ground. The sample air was drawn continuously through a 6 in long ¼ in OD SS tube, which extended ~2 in to the outside bottom of the box and was coupled to two SS crosses, where the flow was diverted to a MoO converter tube for the NOy and a bypass PFA tube of same length for the NO measurement, at 1 slm respectively. All SS components were Teflon coated and temperature controlled at 40 oC. Each flow was filtered by a Teflon membrane filter (Gelman-Teflo) with 2 m pore-size, and directed through a stream selector assembly with mass flow controllers (MFC), all mounted inside the box. The sample residence time inside the PFA tubing between the inlet box on the tower and the CLD unit inside the mobile lab at the ground was ~0.1 s. The chemiluminescence occurred in a gold-plated reaction vessel in front of a single PMT (Hamamatsu R2257) at ~8.5 Torr. The integrated photon counts were recorded at 1 Hz. The instrument background (or “zero”) was measured every 15 minutes for 2 minutes, overlapping the last minute of a NO measure with the first minute of the following NOy measure mode, resulting in a NO and NOy zero respectively. NO and NOy measure modes were switched every 2 minutes. The 1 Hz data were averaged to 1 minutes, and the 1 min zeroes were interpolated over, and subtracted from the 15 min measure mode periods. Automated calibrations were performed via a programmed set of NO, NO2, n-propyl nitrate (NPN), and HNO3 standard additions to the sample inlet on average 3 times per day in ambient air, and once per day in zero air. The calibrations allowed the determination of specific parameters that are relevant for the assessment of the overall instrument performance, such as sensitivity, artifacts, detection limits, and conversion efficiencies of the MoO tube.

The number of calibrations performed in ambient air during each city’s measurement period ranged between 26 and 44. The NO detection limit for a 1 min integration time was ~3 pptv (at S/N =2), for all 3 data sets equally. The average detector NO sensitivity (S_NO) varied between 3.91 ±0.10 and 3.56 ±0.11 Hz/pptv, indicating a NO measurement precision between ±8 and ±13 %. A difference in signal was present when sampling zero air in NO measure mode versus NO zero mode. This NO artifact (A_NO) was largest in Macon with 19 22 pptv when research grade oxygen had to be used instead of zero air; while A_NO was 11 8 pptv in Augusta, and 8 ±5 pptv in Columbus. A_NO was interpolated between calibrations and subtracted from the ambient NO measurements. Since the zero volume efficiency was less than 100 %, i.e. on average between 95 and 98 %, the instrument’s zero varied with ambient NO and NOy levels, respectively. Thus, during low level periods typically occurring at night, the NO_zero signal counts typically averaged 1400 Hz ±1 %. The final NO mixing ratios (ppbv) were determined as follows

NO_ppbv = (NO – NO_zeroipol) / S_NOipol – A_NOipol

with NO being the actual signal counts in NO measure mode.

The accuracy of the NO measurements had uncertainty due to variations in instrument zeroes, sensitivities, MFC calibrations, and the level of calibration standard used. The latter was a compressed, NIST traceable gas tank of 4.58 ±0.09 ppmv NO in O2-free N2 (Scott-Marrin Inc.), which had been cross calibrated prior to the study against 7 other similar standards showing a 2 % deviation from the nominal value as measured by LIF (S. Sandholm pers communication). The MFC calibrations before and after the study were within 2 %. The biggest contributor to the overall uncertainty was the variable level of ambient NO before and after the standard addition and the interpolation necessary for the S_NO determination, which is estimated here at 12 % as reflected by the total S_NO variation. Therefore, the overall uncertainty of the NO measurement is estimated at ±15 % as RMS error of all the above potential inaccuracies.

Each calibration cycle allowed the determination of the instrument’s sensitivity to NO2, NPN, and HNO3. The same NO standard that was used for determining the NO sensitivity was used in a gas phase titration (GPT) cell inside the mobile lab at the bottom of the tower. Titration of NO occurred with O3 produced from zero air in front of a Hg penray lamp, and was set between 98 and 100 % efficiency (TT_eff). The NO2 sensitivity (S_NO2) was calculated by subtracting the signal portion that was due to the untitrated NO cal gas (NO_nontit) from the delta signal increase due to standard addition to the ambient NOy sample (NOy_cal – NOy_amb), divided by the net level of NO2 calibration gas,

S_NO2 = ((NOy_cal – NOy_amb) – NO_nontit) / (TT_eff x [NO_cal]).

The nominal NO mixing ratio resulting from the dynamic dilution process [NO_cal] was 18.8 ppbv. In ambient air, S_NO2 ranged between 1.33 and 2.27 0.55 Hz/pptv revealing a NO2 conversion efficiency Q_NO2 of 36 to 63 ±15 % (= S_NO2 / S_NO x 100). With each calibration cycle the conversion efficiencies for NPN and HNO3 were also determined via standard addition and calculated as follows,

Q_I = (NOy_cal – NOy_amb) / (S_NO x [I_cal]),

with I either NPN or HNO3. NPN cal gas was delivered mass flow controlled to the converter inlet from a NIST traceable compressed air tank of 3.88 ±0.19 ppmv NPN in O2-free N2 (Scott-Marrin Inc.). HNO3 was supplied from a permeation tube (Kin-Tek) inside an oven controlled at 40 ±0.1 oC via a critical orifice controlled zero air flow of ~10 sccm. The permeation rate was verified before and after the study via dissolution of HNO3 using a small scale impinger and subsequent IC analysis of NO3-. The conversion efficiencies for both NPN and HNO3 in ambient air varied between 30 and 60 %, with HNO3 always being 2 to 19 % lower. During the entire study, the conversion efficiency for NO2 was on average 5 % higher than for NPN, with only a few exceptions where they were about the same. Conversion efficiencies for all 3 species were much lower than what had been achieved during the 1999 Atlanta Supersite study (e.g. Q_NO2 = 97 ±7 %). For FAQS 2000 a set of 3 brand new MoO converter tubes from the same manufacturer (Rembar Co., Dobbs Ferry, NY) was used in an alternating fashion, allowing regeneration of one tube while the other one was in operation. Despite frequent regeneration in N2 at 450 to 500 oC, the NO2 conversion efficiency of either one of the 3 tubes never exceeded 82 %, and typically decayed to below 40 % within 2-3 days of sampling. Therefore, the interpolated NO2 sensitivities were used to calculate the final NOy mixing ratios analogous to NO

NOy_ppbv = (NOy – NOy_zeroipol) / S_NO2ipol – A_NOyipol.

The NOy zeroes averaged 2000 Hz 10 %, and an artifact A_NOy was present when sampling zero air. This artifact varied with time and level of converter decay, and was therefore considered in a time-dependent manner; it averaged 0.61 ppbv for Macon, 0.95 ppbv for Augusta, and 0.49 ppbv for Columbus. Based on measured variations in NOy over 2 – 3 h periods, the precision of our NOy measurements ranged between 10 and ±15 %. In addition to the potential uncertainties that contribute to the NO inaccuracies above, our estimate for the overall accuracy of the NOy measurements included the uncertainties of the titration, NO2, NPN, and HNO3 conversion efficiencies resulting in an RMS error of ±25 %.

References

Atlas, E.L., et al., Partitioning and budget of NOy species during the Mauna Loa Observatory Photochemistry Experiment, J. Geophys. Res., 97, 10,449-10,462, 1992.

Bollinger, M.J., Chemiluminescent measurements of the oxides of nitrogen in the clean troposphere and atmospheric chemistry implications, Ph.D. thesis, Chem. Dep., Univ. of Colo., Boulder, 1982.

Fahey, D.W., C.S. Eubank, G. Huebler,, and F.C. Fehsenfeld, Evaluation of a catalytic reduction technique for the measurement of total reactive odd-nitrogen (NOy) in the atmosphere, J. Atmos. Chem., 3, 435-468, 1985.

Fahey, D.W., et al., Reactive nitrogen species in the troposphere: Measurements of NO, NO2, HNO3, particulate nitrate, peroxyacetyl nitrate (PAN), O3, and total reactive odd nitrogen (NOy) at Niwot Ridge, Colorado, J. Geophys. Res., 91, 9781-9793, 1986.

Fehsenfeld, F.C., et al., A study of ozone in the Colorado mountains, J. Atmos. Chem., 1, 87-105, 1983.

Fehsenfeld, F.C., et al., Intercomparison of NO2 measurement techniques, J. Geophys. Res., 95, 3579-3597, 1990.

Kley, D., and M. McFarland, Chemiluminescence detector for NO and NO2, Atmos. Technol.,12, 63-69, 1980.

Parrish, D.D., et al., The total reative oxidized nitrogen levels and their partitioning between the individual speciesat six rural sites in Eastern North America, J. Geophys. Res., 98, 2927-2939, 1993.

Parrish, D.D., J.S. Holloway, and F.C. Fehsenfeld, Routine, continuous measurement of carbon monoxide with parts per billion precision, Environ. Sci. Technol., 28, 1615-1618, 1994.

Ridley, B A and L C Howlett, An instrument for nitric oxide measurements in the stratosphere, Rev. Sci. Instrum., 45, 742-746, 1974.