Techniques For Measuring Reactive Nitrogen Oxides, Volatile Organic Compounds, and Oxidants
A key element in advancing the understanding of tropospheric production of ozone is the ability to make unequivocal measurements of the concentrations of the ozone precursors, the reactive nitrogen oxides (NOy compounds) and volatile organic compounds (VOCs). Because tropospheric chemistry is shaped by oxidants, such as ozone, hydrogen peroxide (H2O2), and oxidizing radicals, the concentration of these compounds also must be measured to test present understanding of atmospheric oxidation mechanisms.
If measurements are to be meaningful, reliable instruments and techniques are necessary. Therefore, it is vital to have trustworthy estimates of the uncertainties in the observations, because these observations are the touchstones against which theoretical understanding is tested. With such estimates, observations and theory can be compared meaningfully; the results from separate studies can be merged reliably; gradients in observations over large distances can be characterized credibly from separate data sets; and time patterns from different monitoring networks can be used to establish longer trend records. In addition, the ability to make accurate measurements of the concentrations of these species, coupled with an adequate theoretical understanding, provides the ability to test whether control measures to regulate the emission of anthropogenic oxidant precursors are effectively curbing the concentrations of these compounds in the atmosphere.
The atmospheric chemistry community has devised a way to address mea-
surement uncertainties that is arduous but effective. It uses formal, rigorous, and unbiased inter- and intramethod comparisons of techniques and instruments. The most instructive of those instrument intercomparisons have the following features:
• Several different techniques are used to measure the same species.
• Insofar as possible, measurements are made at the same place and time and under typical operating conditions.
• Accuracy and precision estimates are stated in advance of the study.
• Atmospheric samples are ''spiked'' with known amounts of species that are potential artifacts. Where possible, samples of ambient air as well as synthetic air are spiked.
• All investigators prepare their results independently and separately from the others.
• Investigators jointly (or through an independent party) compile separate results and assess the state of agreement or disagreement.
• Results and conclusions are published in a peer-reviewed journal.
• The process is repeated occasionally.
There have been several field studies devoted specifically to the assessment of instrument reliability, as opposed to obtaining data to answer a geophysical question (Hoell et al., 1987a,b; Hering et al., 1988; Fehsenfeld et al., 1987, 1990). Because these intercomparisons provide the only objective assessment of instrument capability, the scientific community's knowledge of the accuracy of current instrumentation depends heavily on these results. The instruments and techniques available for measurement of NOy' VOCs, and oxidant species; the basic operating principles of these devices; and highlights of the tests done thus far to determine instrument reliability are summarized below.
Measurement Techniques for Oxides of Nitrogen and Their Oxidation Products
The reactive oxides of nitrogen in the atmosphere are largely nitric oxide (NO) and nitrogen dioxide (NO2), known together as NOx. During the daytime, there is a rapid interconversion of NO and NO2. One important byproduct of this interconversion is the photochemical production of ozone in the troposphere. In addition, NOx is converted to a variety of other organic and inorganic nitrogen species. These are the compounds that make up the reactive nitrogen family, NOy (NOx + organic nitrates + inorganic nitrates).
This chemistry is described in detail in Chapters 5 and 6. The techniques that have been developed to measure these compounds are discussed below.
The reliability of techniques to measure NO has been established rigorously. Two fundamentally different methods have been compared: chemiluminescence (NO-O3 chemical reaction and emission of radiation from the nitrogen dioxide product) and laser-induced fluorescence (absorption of radiation by nitric oxide and then reradiation at different wavelengths by the excited nitric oxide). During two separate tests of these techniques (Hoell et al., 1985, 1987a), chemiluminescence instruments and a laser-induced fluorescence instrument measured ambient concentrations simultaneously at a rural site and from an aircraft. The data agreed within 30% in all of these chemically different environments and over concentrations of NO spanning a range of 0.005 to 0.2 parts per billion (ppb). These results strongly indicate that NO can be measured reliably by either technique under most field conditions.
Many techniques have been developed to measure nitrogen dioxide, but few can measure NO2 at concentrations below parts per billion, and few have been demonstrated to be free of interference from other atmospheric constituents. The standard way to measure NO2 in almost all air quality studies has been to use surface-conversion techniques to convert NO2 to NO and to subsequently detect the NO by chemiluminescence. The conversion techniques include the use of heated catalytic metal surfaces and surfaces coated with ferrous sulfate or other compounds. However, the development of the photolytic NO2-to-NO converter several years ago (Kley and McFarland, 1980) offered a potentially more specific conversion technique, albeit less simple. A recent comparison (Fehsenfeld et al., 1987) made a detailed study of the performance of surface and photolytic methods. In this study, the ferrous sulfate and photolytic converters agreed well at NO2 concentrations of 1 ppb and greater. However, the ferrous sulfate converter systematically reported higher values at lower concentrations, reaching a factor of 2 higher at 0.1 ppb. Spiking tests showed that the ferrous sulfate converter also was converting peroxyacetyl nitrate (PAN) to NO. Hence, whenever PAN is significant in comparison to the NO2, the ferrous sulfate converter gives results that overestimate the concentration of NO2. A heated molybdenum oxide surface converter was
found to convert NO2, PAN, and HNO3 to NO, indicating that heated-surface converters also cannot be considered specific for NO2 or for NOx.
The photolytic converter/chemiluminescence and ferrous sulfate/chemiluminescence techniques were compared during aircraft flights over the eastern Pacific Ocean and the southwestern United States at altitudes of 0.6 to 7.3 kilometers (km) (Ridley et al., 1988). In agreement with the intercomparison discussed above, the ferrous-sulfate-equipped instrument was found to be much less specific for NO2. It registered levels about three times larger than the photolytic converter, presumably because of the conversion of PAN and perhaps other organic nitrates to NO.
Newer technology is emerging to measure NO2. Three techniques that show considerable promise are photofragmentation/2-photon laser induced fluorescence (LIF), tunable-diode laser absorption spectrometry (TDLAS), and luminol chemiluminescence. The LIF and TDLAS techniques provide specific spectroscopic methods to measure NO2; the luminol technique provides a sensitive, portable method with low power requirements. Two recent studies have tested these techniques against the photolysis/chemiluminescence technique (Fehsenfeld et al., 1990; Gregory et al., 1990).
A ground-based comparison (Fehsenfeld et al., 1990) tested the photolysis/ chemiluminescence technique against the TDLAS and luminol techniques. For NO2 concentrations above 0.2 ppb, no interferences were found either for the photolytic converter/chemiluminescence technique or for TDLAS. However, interpretation of the results from TDLAS showed that correlation coefficients should not be used to select the data that are near the detection limit of NO2 for the instrument (Fehsenfeld et al., 1990). At these levels the background noise is normally distributed about the reference NO2 spectrum. Selection of the data with high correlation coefficients would lead to NO2 concentrations that are too high (Fehsenfeld et al., 1990). This test indicated that interferences from PAN and ozone influence the NO2 measurements made using the luminol technique. However, during the comparison those interferences were consistent enough that for NO2 concentrations above 0.3 ppb, they could be corrected using simultaneously measured values of ozone and PAN (Fehsenfeld et al., 1990). Techniques are being developed to remove or separate interfering substances from the ambient air prior to analyses by the luminol detector. However, the effectiveness of these techniques has not been verified by field tests.
An airborne comparison of TDLAS with LIF and photolytic converter/chemiluminescence was conducted by Gregory et al. (1990). The intercomparison of these three instruments in ambient air for NO2 > 0.1 ppb indicated a general level of agreement among the instruments of 30-40%. For NO2 < 0.05 ppb the results indicated that TDLAS overestimates the NO2 mixing
ratio, presumably because of the use of correlation coefficients as the data selection criterion. At these low concentrations, agreement between LIF and photolytic converter/chemiluminescence measurements was within 0.02 ppb with an equal tendency for one to be high or low compared to the other. This 0.02 ppb agreement is typically within the expected uncertainties of the two techniques at NO2 mixing ratios < 0.05 ppb.
It is believed that, properly used, the LIF, TDLAS, and the photolytic converter/chemiluminescence techniques measure NO2 concentrations well below 0.1 ppb, free of significant artifact or interference. These techniques should therefore be able to measure NO2 concentrations throughout the troposphere above North America.
Two instruments, both of which use cryogenically enriched sampling with electron-capture gas chromatography detection, have been intercompared in the remote maritime troposphere (Gregory et al., 1990). At mixing ratios of <0.1 ppb, the two instruments differ on average by 0.017 ppb with a 95% confidence interval of ± 0.009 ppb. At PAN mixing ratios of 0.1-0.3 ppb, the difference between the instruments was 25% ± 6%. A linear regression equation developed by comparing all data < 0.3 ppb from the two techniques gave a line with a slope of 1.34 ± 0.12 and an intercept of 0.0004 ± 0.012 ppb. Although one instrument was consistently high relative to the other for ambient measurements, these levels of agreement were usually within the stated accuracy and precision of the two instruments. These results are reassuring. Nevertheless, their significance is reduced by the similarity in the design and operation of the two instruments.
A test by Hering et al. (1988) focused on the capability to measure nitric acid (HNO3). Over an 8-day period at a site with urban and suburban characteristics, six methods were used to make simultaneous measurements: filter pack, denuder difference, annular denuder, transition flow reactor, tunable-diode laser, and Fourier transform infrared spectrometer. The reported concentrations of HNO3 varied by more than a factor of 2. These differences were substantially larger than the estimated precision of the instruments. The tests indicated that artifacts or interferences exist for some of the sampling methods associated with either the field sampling components (e.g., inlet
lines), the operating procedures, detector specificity, or alteration during sampling in the physical or chemical make-up of the ambient air, such as shifts in the gas- and solid-aqueous-phase equilibrium of HNO3, ammonia, and ammonium nitrates.
Several conclusions could be drawn from Hering's data set. The larger percentage differences in the techniques that were observed at higher HNO3 concentrations and the dependence of the differences on day or night sampling suggest uncontrollable shifts of the equilibrium (ammonium nitrate evaporation) in samples obtained by some instruments. The annular denuder exhibited poor intramethod precision for HNO3, and its average value was substantially below the means of the spectroscopic methods and those of all methods. The results from tungstic acid adsorption tubes and filter packs (> 8-hr sample) deviated substantially from those two means. The falter packs exhibited a positive bias (systematically higher than average HNO3 concentrations) that increased as the sampling time average increased, indicating an artifact due to ammonium nitrate particle evaporation to release HNO3 (and ammonia). The denuder difference, transition-flow reactor, filter pack (<8-hr sample), and spectroscopic methods were in good agreement. This comparison provides a valuable start in assessing the problems of reliable measurement of HNO3.
A more recent test involved three different measurement approaches: nylon falter collection, tungstic oxide denuder, and TDLAS (Gregory et al., 1990). In general, the filter measurements were high relative to those reported by the denuder. No correlation was observed between the falter and denuder techniques for HNO3 < 0.15 ppb. Below 0.3 ppb, the difference between simultaneous measurements from the denuder and filter instruments was greater than the expected accuracy and precision stated for each instrument for more than 75% of the measurements. Comparing the denuder technique and TDLAS, TDLAS measurements were consistently higher; for HNO3 > 0.3 ppb, TDLAS results were systematically higher by a factor of approximately two. There was only one instance of overlap among all three techniques at concentrations of HNO3 well above detection limits. In that case, the measurements from the filter and TDLAS were in agreement, whereas those from the denuder (with only a 35% overlap) were about a factor of two lower. The paucity of simultaneous measurements from all three instruments prevented firm conclusions being drawn from the intercomparison. However, it was clear that there was substantial disagreement among the three techniques, even at mixing ratios well above their respective detection limits. These inter-comparisons clearly indicate that current techniques do not allow the unequivocal determination of HNO3 in the range of concentrations expected in the nonurban atmosphere.
One final comparison is worthy of note. A new technology is emerging to measure HNO3 and other soluble gases. For HNO3, this approach involves the absorption of HNO3 contained in air into ultraclean water followed by analysis with ion chromatography. The approach has been obvious for decades, but only recently have water purification and handling techniques become sufficiently advanced to allow measurement of low levels of these compounds (Cofer et al., 1985). A recent application of this approach is the mist-chamber technique (Cofer et al., 1985; Talbot et al., 1990), which is used to measure nitric acid and other atmospheric acids (Talbot et al., 1988). The mist-chamber technique was recently compared with the nylon Filter method (Talbot et al., 1990). Laboratory and field tests indicated that both techniques were capable of collecting and analyzing HNO3 emitted from permeation tube sources. However, in field measurements made at a rural site, the nylon Filter yielded HNO3 mixing ratios 70% larger than those measured simultaneously by mist-chamber techniqes. Subsequent tests revealed a small positive interference for ozone on the nylon filter, but this interference could not account for the large discrepancy noted above. Talbot et al. (1990) suggested that the nylon filter may suffer interference from other species as well. In any event, this comparison shows the need for caution in interpreting the measurements of HNO3 made with available techniques and underlines the need for further study to determine the reliability of the various methods.
Total Reactive Nitrogen Oxides
Understanding of reaction pathways can be advanced by the measurement of the total abundance of reactive nitrogen compounds, NOy' as well as by the measurement of the individual NOy species. For example, it is NOy' rather than such components as NO2, that is of primary interest in tracking the transport and deposition of tropospheric nitric acid on a regional basis.
Several NOy measurement techniques have been proposed. In general, all rely on the reduction of the NOy-species to NO followed by detection of the NO. A ground-based comparison of two of these techniques, the gold-catalyzed conversion of NOy to NO in the presence of carbon monoxide and the reduction of NO. to NO on a heated molybdenum oxide surface, has been done by Fehsenfeld et al. (1987). The instruments were found to give similar results for the measurement of NOy concentration in ambient air under conditions that varied from clean continental background air to typical urban air, with NOy ranging between 0.4 ppb and 100 ppb. However, it was found that when the molybdenum oxide converter was operated for extended periods (several hours) with NOy concentrations > 100 ppb, the conversion efficiency
dropped significantly. For this reason the gold-catalyzed converter was judged more reliable when used in a polluted environment.
Measurement Techniques for Carbon Monoxide and Volatile Organic Compounds
Unlike the NOx measurement techniques described above, which are adequate, the techniques for measuring VOCs and their oxidation products do not meet current needs. The analysis of VOCs is complicated by the extreme complexity of the mixtures that can be present in the atmosphere. Over one hundred detectable VOCs can be present in air sampled from reasonably isolated rural sites (P.D. Goldan, pers. comm., Aeronomy Laboratory, NOAA, 1990). In urban locations this number is substantially greater (Winer et al., 1989). VOCs emitted by vegetation, estimated to account for 50% of the VOCs emitted into the atmosphere in the United States (Placet et al., 1990), are mainly highly reactive olefinic compounds. Moreover, it is believed that more than one-third of the natural compounds that are emitted are, as yet, unidentified. Air samples obviously can contain many different VOCs of natural and anthropogenic origin; the oxidation of each of these species creates a mixture that contains many additional oxidation products as well. This complex chemistry is discussed in detail in Chapters 5 and 6. It is clear that the analysis of VOCs and their oxidation products is a formidable task. The techniques that have been developed to measure these compounds, as well as several promising new methods, are described in the following sections.
The discussion of these measurement techniques is divided into five parts: (1) carbon monoxide (CO); (2) nonmethane hydrocarbons (NMHC), which are compounds other than methane (CH4) that are composed entirely of hydrogen and carbon; (3) formaldehyde (HCHO), the simplest aldehyde; (4) other aldehydes and ketones; and (5) organic acids. These are the compounds that most strongly influence the photochemical production of tropospheric ozone, which have the most atmospheric variability (as opposed to CH4), and for which measurement capability is a matter of the most concern. The term "volatile organic compound" refers to all the above compounds except CO, and also refers to other compounds, such as organic nitrates, peroxides, and radicals, that are discussed elsewhere in the report.
CO is ubiquitous in the atmosphere, and it has many sources, both natural
(oxidation of methane and other natural VOCs and biomass burning) and anthropogenic (combustion processes). Its lifetime is long enough, on the order of a few months, that it is distributed globally, and its concentration ranges from 50 to 150 ppb in the remote troposphere. Concentrations above this background can indicate air masses that have had recent anthropogenic pollution input. Given the relatively unreactive nature of the gas and its high concentrations, it is expected to be one of the more easily measured trace atmospheric species.
Three techniques are used widely for measurement of CO in the troposphere: collection of grab samples followed by analysis using gas chromatography (GC); tunable-diode laser absorption spectrometry (TDLAS); and gas filter correlation, nondispersive infrared absorption spectroscopy (NDIR). The first two of these methods have been compared in rigorous sets of tests (Hoell et al., 1987b, and the references therein). In the earlier ground-based comparison of three GC techniques (one with direct injection of samples) and one TDLAS system, there was a high degree of correlation between the results of all four techniques in ambient measurements and for prepared mixtures of CO in ambient air (Hoell et al., 1987b). The general level of agreement was within 15%. However, a day-to-day bias between the techniques was observed to result in differences between techniques as large as 38%.
In the later airborne comparison (involving the two grab sample GC and the TDLAS systems), the techniques had been refined to the point that, at mixing ratios of 60-140 ppb, the level of agreement observed for the ensemble of measurements was well within the overall accuracy stated for each instrument (Hoell et al., 1987b). The correlation coefficient determined from the measurements taken from respective pairs of instruments ranged from 0.85 to 0.98, with no evidence of the presence of either a constant or proportional bias between any of the instruments. Thus, the reliability of the measurement of CO has been rigorously established.
Much of the reactive carbon entering the atmosphere is in the form of nonmethane hydrocarbons (NMHCs). The standard approach for measurement of NMHCs is based on GC separation of individual hydrocarbons and the detection of each using a flame ionization detector (FID). Singh (1980) summarized the general procedures used in the analysis of ambient hydrocarbon samples. The GC column and temperature programming of the column are selected to give the desired resolution of the compound peaks. Over the years, separation and the integrity of compounds that pass through the columns have been improved by development of better column packing compounds for packed columns or coatings for open tubular columns.
The FID is a nonspecific hydrocarbon detector with a sensitivity that, in general, is linearly proportional to the number of carbon atoms in a VOC molecule (Ackman, 1968). Compound identification is usually established by compound retention in the GC column. In addition, mass spectrometric identification of given peaks can be made to confirm the elution time assignments or can be used to help in the identification of an unknown peak.
When very low concentrations of the VOCs are measured, it is necessary to concentrate the samples before they are injected into the column. This is done by concentrating a volume of air cryogenically or with a trapping matrix before injection onto the column. By temperature programming the trap, it is possible to separate the VOC compounds to be measured from compounds that comprise the bulk of the air sample: nitrogen, oxygen, water vapor, argon, and carbon dioxide. As a consequence of these improvements, VOCs with concentrations as low as 5 ppt in air have been measured with good resolution by GC-FID systems. However, there can be problems with using this method to measure reactive VOCs at low concentrations in air. Large amounts of compounds, particularly high-carbon-number compounds, can be retained by the trapping medium. In addition, reactions between the VOCs and oxidants, such as residual ozone that survives the collection procedures, may destroy some hydrocarbons and produce other compounds not originally in the sampled air. Hence, additional methods are required to reduce the oxidants to negligible concentrations before preconcentration without altering the concentration of the hydrocarbons to be analyzed.
Often, measurements of NMHCs in the field are done under circumstances that require maximum portability and low power consumption, and they arc done in an environment adverse to the operation of a sensitive instrument. Consequently, many NMHC measurements are done by acquiring an air sample in a suitably prepared container and transferring it into the GC-FID. Sample containers have been made from glass, treated metal, and special plastics. Sampling procedures often require that the containers be purged before the air sample is obtained and that the sample be stored in the container above atmospheric pressure. Although much has been done to ensure the integrity of the compounds of interest in these containers, many of the difficulties attendant to this approach are associated with the stability of the sample during transport and storage. When samples of hydrocarbons are analyzed after having been stored for several days, substantial losses of the heavier hydrocarbons can occur (Holdren et al., 1979). This is significant because occasionally, several months may pass before hydrocarbon samples stored in these containers can be analyzed.
Thus far, there has been no rigorous comparison of the various versions of GC-FID systems and sampling containers. However, the limited comparisons that have been done indicate that the techniques could be satisfactory for
measuring relatively high concentrations (> 10 ppb carbons) of simple, light hydrocarbon compounds (five carbons or fewer).
The oxidation of NMHC forms the carbonyl compounds, the aldehydes and ketones (CHO). Measurements of these compounds can test present understanding of the VOC oxidation mechanism. The oxidation of aldehydes and ketones can be an additional source of ozone and oxidizing free radicals, and the photolysis of aldehydes and ketones can be a primary source of radicals. The simplest aldehyde, formaldehyde (HCHO), is particularly important because it can be formed by the oxidation of methane. As a result, it is distributed throughout the troposphere. HCHO can also be emitted into the atmosphere as a direct product of hydrocarbon combustion (Lawson et al., 1990a). Thus the photolysis of HCHO could be a key process in the formation of tropospheric ozone.
Four techniques have emerged for the measurement of HCHO: TDLAS; enzymatic fluorometry (EF), which involves the absorption of HCHO from a sampled air stream into water followed by detection of the fluorescence from the reaction of the aqueous HCHO with b-nicotinamide adenine dinucleotide, as catalyzed by the enzyme formaldehyde dehydrogenase; a diffusion scrubbing fluorescence (DSF) technique, which involves the absorption of HCHO from a sampled air stream into water followed by detection of the fluorescence from the reaction of the aqueous HCHO with ammonia and acetylacetone; and a derivatization technique, which involves trapping HCHO on a substrate impregnated with 2,4-dinitrophenylhydrazine (DNPH) followed by extraction of the derivatized compounds and ultraviolet absorption analysis. Two studies have been reported that compare these techniques for the measurement of HCHO in ambient air. In the first, Kleindienst et al. (1988) compared the four techniques for the measurement of HCHO at the lower concentrations (< 10 ppb) typically found in rural air. Because of its recent development, potential interferences for the DSF technique were not known in advance of the study and, as a consequence, the DSF technique was not involved in the ambient air measurements. In this study, no large systematic errors were observed in synthetic air mixtures with and without added interferants such as NO2, SO2, O3, and H2O2, or, for the TDLAS, EF, and DNPH techniques, in ambient air where ambient concentrations of HCHO ranged from I to 10 ppb. Although reasonably low concentrations of HCHO were encountered during this comparison, no attempt was made to establish detection limits for these instruments.
In a more recent comparison, Lawson et al. (1990a) evaluated the four
instruments in a reasonably polluted urban environment. In this evaluation two additional techniques were included, Fourier transform infrared spectrometry (FT-IR) and differential optical absorption spectrometry (DOAS). Because of their low sensitivity, these latter two techniques are not suitable for the measurement of HCHO in the nonurban atmosphere. However, because they are highly specific optical techniques that can measure higher concentrations of HCHO over limited paths in the free atmosphere, in this urban environment they provided independent measurements for comparison with the measurements made by the other techniques.
During the course of the 10-day study, a systematic diurnal variation was observed in the HCHO; it reached a maximum during the day and a minimum during the night. The average hourly ambient HCHO ranged from 4 to 20 ppb. Because reasonably high concentrations of HCHO were observed during the early morning rush hour, it was surmised that formaldehyde was being emitted directly into the atmosphere as a primary pollutant. Over the study period, the three spectroscopic techniques agreed to within 15% of the mean of these three methods. DNPH yielded values 15-20% lower than the mean of the spectroscopic techniques, whereas DSF yielded values 25% lower than the mean. Measurements obtained with the EF were found to be 25% higher than the mean. Measurements reported early in the study for DSF and EF were closer to the spectroscopic mean; problems developed in these instruments as the comparison progressed. The slight negative bias in the values obtained with DNPH was tentatively attributed to a negative ozone interference (ozone concentrations ranged from 0 to 240 ppb in this field study).
Other Aldehydes and Ketones
The measurement of higher molecular weight aldehydes and ketones has been performed with two different techniques, DNPH cartridges and GC-FID. DNPH has been the standard method for most field measurements of the carbonyls. For these compounds, the method has proven to have adequate selectivity. However, it suffers from low resolution and sensitivity when compared to GC-FID. Also, because DNPH involves liquid extraction of the compounds of interest from the cartridge, blank levels are a problem for measurements of carbonyl compounds at concentrations expected in the rural environment. GC-FID offers reasonable sensitivity and high resolution when capillary columns are used. This technique can achieve detection limits of <0.01-0.2 ppb in one liter of air, depending on the compound analyzed. In GC-FID analysis of ambient air, artifact formation of carbonyl compounds can arise in the cryogenic collection of an air sample. Thus far there have been
no intercomparisons of these techniques for the measurement of the aldehydes and ketones.
Although organic acids can be major components of atmospheric acidity, few measurements of these acids in the vapor phase have been reported (Norton, 1987; Farmer et al., 1987). The mechanisms for their formation are not well understood, and their role in tropospheric chemistry is uncertain. Organic acids can be removed from the atmosphere by deposition. Not enough is known about their chemistry and atmospheric distribution to predict how their oxidation will influence ozone formation.
There are many measurement techniques for collecting organic acids, but few tests have assessed their validity. Keene et al. (1986) compared techniques for collecting aerosol- and vapor-phase formic and acetic acid. The acids were collected in mist chambers, as cold plate condensates, in resin cartridges, in sodium-hydroxide-coated denuder tubes, in sodium-hydroxide-impregnated glass filters, in nylon filters, and on cellulose fibers impregnated with sodium or potassium. The study was limited to ambient air sampling. After collection, all of the samples were analyzed by ion chromatography. The mist chamber and denuder tube gave results that were statistically indistinguishable. The cold plate technique gave results that were in general agreement with the mist and denuder techniques but showed significant differences on some occasions. The nylon filters were found not to retain the acid vapors quantitatively. The sodium carbonate filters gave concentrations somewhat below those of the mist and denuder techniques. The resin cartridges, sodium-hydroxide impregnated glass filters, and the sodium- and potassium-impregnated cellulose filters gave concentrations substantially larger than those of the mist chamber and denuder tubes. Although this study was not able to establish a generally reliable method to measure organic acids, several correctable problems were identified with the techniques. The conclusion was that strong-base-coated filters and GC resin techniques suffer from interferences that can cause serious overestimation of the concentration of organic acids in the air samples.
Measurement Techniques for Oxidants
The oxidants discussed in this section are the hydroxyl radical (OH), the peroxy radical (HO2), ozone, and hydrogen peroxide (H2O2). The processes
responsible for the formation and loss of these highly reactive compounds and the roles these oxidants play in atmospheric chemistry are described in Chapters 5 and 6. The techniques that have been developed to measure these oxidants are described in the following sections.
The Hydroxyl Radical
Atmospheric oxidation is thought to be initiated by OH. However, to date, the verification of this chemistry has been derived solely from laboratory studies and from computer modeling of the chemistry. Neither the concentration of OH nor that of HO2 has been measured in the atmosphere with instruments of established reliability to demonstrate whether the current mechanistic understanding of these fundamental processes is correct.
The requirements for such instruments are challenging. Although huge quantities of OH are generated during the sunlit hours by the photolysis of ozone, the very high reactivity of these oxidizing radicals implies that their atmospheric concentrations are small, typically less than 107/cm3 (˜ 0.4 ppt) (Crosley and Hoell, 1985). Moreover, these free radicals can be lost by collision with the surfaces of instrumentsfor example, inside sampling inlets. Hence, although substantial effort has already been invested in the development of methods to measure OH, a definitive measurement in the troposphere is still to be done (Hoell, 1983; Crosley and Hoell, 1985; Platt et al., 1988; Smith and Crosley, 1990; Armerding et al., 1990; Hofzumahaus et al., 1990a). Techniques currently under development include in situ methods that use laser-induced fluorescence (Davis et al., 1981; Wang et al., 1981; Rodgers et al., 1985; Hard et al., 1986; Chan et al., 1990), a radioactive tracer technique (Campbell et al., 1986; Felton et al., 1990), long-path, laser absorption methods (Huebler et al., 1984; Perner et al., 1987; Dorn et al., 1988; Platt et al., 1988; Hofzumahaus et al., 1990b), and ion-assisted OH detection (Eisele and Tanner, 1990).
Peroxy and Organic Peroxy Radicals
Tropospheric measurements of peroxy (HO2) and organic peroxy (RO2) free radicals have been made with two different techniques, peroxy radical chemical amplification (PeRCA) and matrix isolation with electron spin resonance detection (MIESR). MIESR (Volz et al., 1988) relies on the cryogenic trapping' of HO2 radicals in a water matrix followed by the detection of the free radical using electron spin resonance. Problems with interference in the
ESR spectra have been overcome by using deuterium oxide (D2O), instead of water, as the isolation matrix. This substitution has improved the signal-to-noise ratio and spectral resolution, allowing the identification of different free radical species during field measurements (Volz et al., 1988). The PeRCA technique relies on the oxidizing ability of the odd-hydrogen free radicals to convert NO and CO to NO2 and CO2 in a chain reaction (Cantrell and Stedman, 1982; Cantrell et al., 1984). The NO2 produced in the chain reaction is measured using luminol chemiluminescence. Measurements of ambient concentrations of HO2 free radicals in the atmosphere using this technique have also been reported (Cantrell et al., 1988).
Both techniques claim detection limits for HO2 on the order of 1 ppt, which is sensitive enough for most ambient measurements. PeRCA has the distinct advantage of providing a continuous record of the total HO2 radical concentration in an air mass. MIESR, on the other hand, provides an integrated measure of the HO2 radical concentration, but it has the advantage of speciation. Calibration of both techniques under ambient conditions remains a research challenge. There have been no formal intercomparisons of these techniques, and their ability to measure HO2 radical reliably is open to question.
In addition to these techniques specifically designed to detect HO2, many of the in situ OH measurement techniques outlined in the preceeding section might be adapted to measure HO2. Such adaptation requires that the HO2 in the ambient air sampled by the instrument be titrated to OH, probably by reaction with NO before detection.
Over the years several techniques have been developed to measure ozone. These include absorption of ultraviolet (UV) light, chemiluminescence, and chemical titration methods, particularly electrochemical techniques. Each has advantages for certain kinds of tropospheric ozone measurements.
The absorption of UV light by the ozone molecule provides a reasonably straightforward and accurate means to measure ozone. Most instruments rely on the 254 nm (nanometer) emission line of mercury (which happens to coin-tide with an absorption maximum of ozone) from a mercury discharge lamp as the UV light source. This technique, which has been incorporated into several high-quality commercially available instruments, is reliable, and interference that occurs because of the absorption of the UV light by molecules other than ozone can generally be ignored. Most high-quality, routine in situ measurements of ozone have been made with this technique.
The chemiluminescence produced by the reaction of ozone with nitric oxide forms the basis for a sensitive and specific ozone detection method. (The reactions of ozone with an unsaturated NMHC such as ethylene also have been used, but they are somewhat less sensitive.) Although chemiluminescence tends to be more complicated than the UV absorption method, it can make fast-response ozone measurements because of its greater sensitivity. For this reason chemiluminescence has been used to measure ozone fluxes that can be deduced from the correlation of ozone variation with atmospheric turbulence.
Electrochemical sondes measure the electrical conductivity of an electrolytic solution and rely on the conversion of chemicals in the solution by ozone in the sampled air, which alters the conductivity of the solution. A typical instrument, such as the electrochemical concentration cell (Komhyr, 1969), is composed of platinum electrodes immersed in neutral buffered potassium iodide solutions of different concentrations in anode and cathode chambers. When ozone-containing air is pumped into the cathode region of the cells, a current is generated proportional to the ozone flux through the cell. Sondes can be very lightweight and can therefore be lifted by small balloons; they are generally used to measure ozone profiles in the atmosphere. However, the measurements made by these instruments can suffer from positive or negative interference by compounds other than ozone, and this decreases the reliability of the instrument (Barnes et al., 1985).
Over the years, several formal and informal intercomparisons of these techniques have been made (Attmannspacher and Dutsch, 1981; Aimedieu et al., 1983; Robbins, 1983; Hilsenrath et al., 1986). Although the most recent and comprehensive of these studies was aimed at evaluating instruments used to measure stratospheric ozone, many of the findings can be applied to tropospheric ozone. The consensus to be drawn from these comparisons indicates that the best UV absorption instruments are probably reliable for measurement of tropospheric ozone with uncertainties of less than 3%. Chemiluminescence instruments should be equally good. The electrochemical sondes are susceptible to interference that reduces their intrinsic accuracy somewhat. However, it must be emphasized that all of these techniques when used in routine measurement likely will be subject to much larger uncertainties.
Several techniques are used to measure H2O2. These techniques have been subjected to an urban-area comparison (Kleindienst et al., 1988), and a second comparison is under way (Sakugawa and Kaplan, 1990). Kleindienst et al.
(1988) compared four techniques used to measure H2O2 in air: tunable diode laser spectrometry using infrared absorption to measure H2O2 (Slemr et al., 1986); continuous-scrubbing extracting gas-phase peroxides into aqueous solution that are analyzed by enzymatic fluorometry (Lazrus et al., 1986); diffusion-scrubbing followed by analysis using enzymatic fluorometry (Hwang and Dasgupta, 1986); and continuous-scrubbing extracting H2O2 into a luminol solution where it undergoes a chemiluminescence reaction (Zika and Saltzman, 1982).
The first three of these techniques were compared for the measurement of H2O2 (1) in synthetic air, sometimes spiked with common interferences; (2) in synthetic air containing UV irradiated mixtures of NMHC and NOx; and (3) in ambient air. The luminol technique was not included in the last phase of the comparison; cf. Kleindienst et al. (1988). For the comparisons done in synthetic air and ambient air, the agreement was satisfactory-30% or better for the three techniques evaluated. These three techniques had detection limits for the measurement of H2O2 of approximately 0.1 ppb. In the tests done in synthetic air containing irradiated mixtures of NMHC and NOx, the agreement among the techniques was not as good, suggesting the presence of some as-yet unidentified H2O2 interference. In this regard, one current concern is the degree to which H2O2 is contained on aerosols. It is not known how much of this aerosol H2O2 is measured by the various techniques.
The techniques using enzymatic fluorometry can also be used to measure organic peroxides. However, these techniques have yet to be tested through intercomparison.
Condensed-Phase Measurement Techniques
The discussion to this point has centered on techniques to study gas-phase chemistry. Current understanding suggests that the production of ozone from precursors during the summer principally involves gas-phase processes. However, the formation and removal of ozone, particularly in seasons other than summer, may involve the condensed phase, including aerosols, fog droplets, or cloud droplets, and may depend on the chemistry that occurs in this phase. During the past 5 years, the potential importance of heterogeneous chemistry (chemistry occurring in more than one phase) has been demonstrated by research aimed at understanding the suppression of stratospheric ozone in polar regions in early spring (Geophys. Res. Lett., 1990).
There are review articles that discuss the role of multiphase chemistry in the troposphere (e.g., Charlson et al., 1991) and the techniques used to study these processes (e.g., Simoneit, 1986; Bidleman, 1988; Ayers and Gillett, 1990;
Turpin and Huntzicker, 1991). However, a review of the techniques that contribute to the measurement of condensed-phase processes as they pertain to the production and destruction of ozone in the troposphere is beyond the scope of this chapter. Inferences made from measurements using these techniques, as they pertain to the distribution of tropospheric ozone, would be premature. Nevertheless, there is evidence that the condensed phase may contain significant quantities of ozone precursors such as the NMHC (Simoneit, 1986; Sicre et al., 1987; Foreman and Bidleman, 1990; Pickle et al., 1990; Brorström-Lundén and Lövblad, 1991; Mylonas et al., 1991; Simoneit et al., 1991). The development of measurement techniques to study multiphase chemistry in the troposphere, the validation of these techniques, and the application of these techniques to atmospheric measurements should be encouraged.
Long-Term Monitoring and Intensive Field Measurement Programs
The measurement techniques described here must be used in well-designed field studies to collect the data necessary to evaluate tropospheric ozone production. Particularly fruitful field studies have typically fallen into two categories: long-term monitoring of one or a few easily measured species and short-term intensive field campaigns measuring a wide suite of atmospheric species involved in chemical transformations. As a guide to planning future field studies, it is worthwhile to consider these two categories.
Long-Term Monitoring Programs
This approach is exemplified by the measurements of CO2 that have been carried out continuously since 1957 to determine the global CO2 background. These measurements clearly have established the increasing trend of CO2 in the atmosphere and have defined its seasonal cycle. More recently, monitoring of ozone regionally and in specific locations has been instituted. The data set is still too short-term to reveal unambiguous trends in urban, suburban, or rural areas. Ozone precursors have not been monitored consistently over large areas, in part because of the difficulty and expense of such measurements. If resources can be found, it is realistic to begin such programs, because suitable instrumentation is now available for monitoring NOx, NOy, and NMHC. The use of atmospheric measurements to determine temporal trends in emissions will provide valuable checks for estimated emission reductions that inventories purport to show.
Designing the long-term monitoring programs will require careful thought. CO and ozone can be monitored with commercially available instruments. Because of the relatively long atmospheric lifetimes of these species, continuous measurements in appropriate locations can provide the necessary trend data. Investigators can rely on mixing in the atmosphere to provide representative samples. However, the important precursors of ozone are quite short-lived, and they undergo rapid chemical transformations in the atmosphere. Measurements at any location will reflect the intensity of proximate and distant sources, atmospheric mixing and transport, and the degree of chemical transformation. Hence, extraction of trends in emissions could be difficult.
The proper choice of sampling location and season could ease these difficulties. Measurements in late fall in suburban areas could be most fruitful because this season often provides low sunlight, temperatures, and precipitation, which will slow chemical transformations. Also during this season the lower troposphere is most stable, which will reduce the variability in atmospheric mixing. A suburban site could provide the best compromise between long transport times, which increases variability in degree of chemical transformation, and poor mixing due to proximity of sources, which increases variability due to specific source input. It is clear that monitoring of trends in emissions cannot easily be a part of a program designed to measure atmospheric transformation processes, because one obscures the other.
The simplest plan would be to concentrate on short-term, intense field studies in which the measurements are done by intercomparison-validated, high-quality instruments. The measurements could be made during one, or at most, a few periods each year rather than as an extended, routine program of measurements done throughout the year. It should be possible to adequately characterize the meteorology of each period to reduce the variability associated with varying transport processes. In this connection, it would also be fruitful to look at ratios of the species concentrations, because the ratios are much more independent of atmospheric mixing processes than are the concentrations themselves.
Intensive Field Studies
Understanding of ozone production outside urban areas has been greatly advanced by several field studies designed to elucidate atmospheric photochemical processes (for example, Dennis et al., 1990). The experience from these studies is that much more is learned from the simultaneous measurement of many photochemical processes and meteorological events than is possible from the separate measurement of each.
Intensive field studies have evolved to cover a wide range of the relevant variables. Two such studies, one carried out in an urban environment and the other in a rural location, are discussed by way of example.
A comprehensive urban study (Lawson, 1990 and the references cited therein), the Southern California Air Quality Study (SCAQS), was carried out in the South Coast Air Basin (SoCAB), which experiences the most severe air pollution in the United States. The SCAQS study was aimed at obtaining an integrated, basin-wide data base of the most important species contributing to air pollution in Los Angeles. By clarifying the VOC/NOx/ozone relationships, these data will aid in improving the models used to design attainment strategies for ozone. The study, which was carried out in 1987, consisted of an intensive 11-day period in the summer and a second 6-day intensive period in the fall. The measurement suite included two highly instrumented ground sites (designated class ''A'' sites), 9 less instrumented support sites (class "B" sites), and 36 sites that make the routine measurements within the framework of the ongoing SCAQ Management District (SCAQMD).
The measurements made at the sites included surface meteorological parameters (temperature, dew point, wind speed and direction); gaseous species (H2O2, NOx, O3, CO, SO2, HNO2, and NO3); and organic vapors (formaldehyde, RO radicals, organic acids, aldehydes, ketones, alcohol's, C1-C12 hydrocarbons, and PAN). Continuous measurements of aerosols included: particle sulfur, particulate matter, sulfate, organic carbon, elemental carbon and black carbon. Aerosol chemistry sampling included: mutagens, metals, organics, particulate matter, carbon, NO3-, SO4-, and polyaromatic hydrocarbons. Size resolved aerosol chemistry involved organic carbon, elemental carbon, and nighttime bromine and lead. These surface measurements were augmented by rawinsonde measurements, and the summer study was augmented by aircraft measurements.
The preliminary results of the study were presented at a symposium of the Air and Waste Management Association in 1989. Several tentative objectives and conclusions have been drawn for each phase of the study. Emissions inventories were checked using measurements capable of testing the reliability of emission factors used to generate the emissions inventories of NOx, CO, and NMHC for point and area sources that are presently used in model calculations. It was found from tunnel studies that the emission factors being used to generate inventories for CO and NMHC for mobile sources may be too small. Also it was recognized that better inventories for ammonia emissions are required, because the formation of ammonium nitrate may be a significant source of aerosol formation and play a significant role in atmospheric denitrification. Transport was elucidated using tracer studies to determine: (1), in the summer, the relative impacts of elevated and ground level sources on
the concentration of ozone and NO2 during on- and off-shore flow; and (2), in the fall, the relative influence of elevated and ground level emission sources of NOx on the ground level NOx during stagnation events. Measurements of gas-phase pollutants and photochemistry elucidated the role of NOx and NMHC in the formation of ozone and other oxidants.
The extensive data base obtained from this study has been archived and is being analyzed by various investigators. A primary goal of the SCAQS is to make this data base available to the scientific community with the aim of developing better models to simulate the atmospheric composition in urban areas.
A recent study (Parrish, 1990), Rural Ozone in a Southern Environment (ROSE), was organized at a ground site in Alabama. This study was part of the Southeastern Regional Oxidant Network (SERON) of the Southern Oxidants Study (SOS). Near the surface the chemical species measured included ozone and other oxidants (H2O2, organic peroxides, and HO2 radicals), the major ozone precursors (NO, NO2, and NMHC), their intermediate oxidation products (organic nitrates, HNO3, NOy, aldehydes, ketones, organic acids, and organic aerosols) and other primary pollutants (CO, SO2, and sulfate aerosols). Meteorological parameters (e.g., temperature, relative humidity, wind speed and direction) also were measured at the surface. The concentration measurements were supported by measurements of biogenic emissions of NOx and NMHC and of surface fluxes of CO2 and ozone. The data from this surface site were supplemented by regional measurements of ozone, NO, NO2, NMHC, organic nitrates, and NO. from an aircraft. The vertical distributions of ozone and NMHC above the site were measured by the aircraft and by tethered balloons. The evolution of the planetary boundary layer, which controls much of the mixing in the lower troposphere, was monitored by balloon-launched radiosondes and by wind measurements by SODAR and boundary layer radar systems.
Other integrated, intensive field studies now planned or under way include the San Joaquin Valley Air Quality Study (SJVAOS)/Atmospheric Utility Signatures, Predictions, and Experiments (AUSPEX) (Roth, 1988; Ranzieri and Thuillier, 1990); the Lake Michigan Ozone Study (Bowne et al., 1990); and the Southern Oxidants Study (Chameides and Rogers, 1988). Nearly all present understanding of ozone formation in rural and remote environments has come from such integrated, intensive field studies, but much remains to be done. Previous studies have been limited in spatial coverage; sites in a wide variety of areas must be studied. Studies have been carried out only in the summer; future studies must compare atmospheric processes between seasons. The free radicals that drive atmospheric chemistry must be measured directly. Current measurements and computer models almost exclusively
address homogeneous, gas-phase chemistry; the role of aerosols must also be investigated. These last two needs will require advances in instruments and in techniques.
In the planning and execution of field studies, chemical-dynamic models are useful for designing the measurement strategy, not just in interpreting the measurement data. In this way modelers can help identify crucial site characteristics, choose critical species and parameters to measure, and optimize the measurement schedule. In addition, the results of the field studies can best provide critical input to the modeling community that will lead to the improvement of computer models.
Reliable techniques for measuring NOx (oxides of nitrogen), even at the low concentrations found in rural and remote air, have been developed recently, and methods that measure nitric oxide (NO), nitrogen dioxide (NO2), peroxyacetyl nitrate (PAN), and the total concentration of the nitrogen oxide (NOy) have been compared These intercomparisons indicate that methods are now available to measure these species throughout the troposphere. However, the validated techniques are of recent origin. In particular, the instruments used to measure NO2 in most air-quality studies usually rely on heated surface converters to transform NO2 to NO. All these methods will likely convert other NOy species to NO as well, and hence can be subject to significant interference. Intercomparison of techniques used to measure HNO3 shows significant variations among methods, and no definitive conclusions can be drawn about the reliability of individual techniques. Further development of HNO3 measurement techniques is required before measurements of that compound can be considered reliable.
For VOCs, only the techniques that measure CO can be considered fully reliable. There has been no intercomparison of methods to measure non-methane hydrocarbons (NMHCs), so there is no way to judge the quality of the large body of data obtained using those techniques. As a general rule, however, measurements made at low NMHC concentrations (atmospheric mixing ratios < 1 ppb of the compound) must be considered suspect. In addition, the heavier NMHCs (C5 and larger) are subject to larger sampling uncertainties than are the lighter NMHCs. Sampling techniques for partially oxidized NMHCscarbonyl compounds, including formaldehyde and organic acidsare under development. However, the data base of measured concentrations of those compounds using this emerging technology is limited.
The basic test of our understanding of oxidizing properties awaits the fur-
ther development of techniques to measure the oxidizing radicals, hydroxyl radical (OH) and peroxy radical (HO2). Much progress has been made toward the development of the necessary instrumentation, but reliable measurements are not yet available.
Reliable techniques to measure ozone are available, and the large data base obtained using these techniques is a valuable resource. However, the uncertainty of routine measurements using these techniques may be as much as 10%, which could limit the conclusions that can be drawn from the data. Intercomparisons have been made for techniques that measure H2O2. There is qualitative agreement among the techniques, but the data obtained on atmospheric concentrations of H2O2 are limited.
Much still must be learned about instrument reliability. An essential factor in establishing reliability is the existence of two or more field-worthy techniques that measure the species of interest. In addition, reliable or standardized calibration procedures must be available for these species. Even with those necessary conditions, the road to harmony can be long and twisting. For example, there is no one recipe for what to do when two or more methods disagree significantly.
Although it is arduous, time-consuming, and costly to develop individual instruments and track down discrepancies, it should be recognized that multiple techniques are essential (and are not wasteful duplication) and that inter-comparisons are vital (and are indeed as much a part of doing atmospheric science as is gathering data to test a geophysical hypothesis). Without inter-comparisons among different techniques, there is no assurance that what is measured is indeed correct.
Accurate and precise measurements of the trace species involved in ozone chemistry are needed to advance the understanding of the formation of high concentrations of ozone, to verify estimates of precursor emissions, and to assess the effectiveness of ozone control efforts. However, these species have not been adequately monitored. As a result, it is not known whether the lack of success of ozone control efforts is the result of failure to achieve targeted reductions in ozone precursors or failure to set appropriate targets. Also, questions remain about the relative importance of anthropogenic and biogenic VOCs, the extent to which ozone production is VOC-limited or NOx-limited, and the role of VOC and NOx oxidation products in ozone formation.
To answer these questions, it is necessary to have reliable measurements of ozone, NOx, VOCs, CO, and the oxidants that catalyze ozone production. Although reliable techniques for measuring many of these species have been available for several years, most of the data bases discussed in this report were not obtained using such techniques. Moreover, measurements made by inexperienced operators using sophisticated techniques may contain uncertainties
that could mask important trends. Only measurements made by skilled operators with reliable instruments can ensure that the science on which emission controls are based is correct and that the effectiveness of these controls is adequately assessed.