Acetone in the atmosphere: Distribution, sources, and sinks

Acetone (CH3COCH3) was found to be the dominant nonmethane organic species present in the atmosphere sampled primarily over eastern Canada (0–6 km, 35°–65°N) during ABLE3B (July to August 1990). A concentration range of 357 to 2310 ppt (= 10−12 v/v) with a mean value of 1140±413 ppt was measured. Under extremely clean conditions, generally involving Arctic flows, lowest (background) mixing ratios of 550±100 ppt were present in much of the troposphere studied. Correlations between atmospheric mixing ratios of acetone and select species such as C2H2, CO, C3H8, C2C14 and isoprene provided important clues to its possible sources and to the causes of its atmospheric variability. Biomass burning as a source of acetone has been identified for the first time. By using atmospheric data and three-dimensional photochemical models, a global acetone source of 40–60 Tg (= 1012 g)/yr is estimated to be present. Secondary formation from the atmospheric oxidation of precursor hydrocarbons (principally propane, isobutane, and isobutene) provides the single largest source (51%). The remainder is attributable to biomass burning (26%), direct biogenic emissions (21%), and primary anthropogenic emissions (3%). Atmospheric removal of acetone is estimated to be due to photolysis (64%), reaction with OH radicals (24%), and deposition (12%). Model calculations also suggest that acetone photolysis contributed significantly to PAN formation (100–200 ppt) in the middle and upper troposphere of the sampled region and may be important globally. While the source-sink equation appears to be roughly balanced, much more atmospheric and source data, especially from the southern hemisphere, are needed to reliably quantify the atmospheric budget of acetone.


INTRODUCTION
Natural and anthropogenic nonmethane hydrocarbons (NMHCs) are released into the atmosphere at a rate of nearly 103 Tg/yr [Duce et al., 1983;Singh and Zimmerman, 1992]. (NAS, 1976;Lloyd, 1979;Atkinson, 1990]. There is further evidence that direct emissions of carbonyl species from both natural and manmade sources are also quite common [NAS, 1976;Isidorov et al., 1985;Sigsby et al., 1987]. Carbonyl compounds are of interest to atmospheric chemists because of their potential toxicity, their ability to photolyze and produce free radicals, their ability to form stable atmospheric products, and their interactions in the smog cycles. Because they are frequently intermediate products of atmospheric oxidation, these molecules can serve as excellent tracers for the validation of photochemical models. While several carbonyl species have been studied in polluted atmospheres [Vairavamurthy et al., 1992], formaldehyde is the only one to receive significant attention in global tropospheric chemistry studies. One of the most abundant reactive oxygenated species in the remote atmosphere is acetone. It was first measured by Cavanagh et a l. [1969] at a concentration of about 1 ppb in the uncontaminated Arctic air of Point Barrow, Alaska. S ingh and Hanst [1981] used a photochemical model to suggest that acetone is an ubiquitous atmospheric species resulting from the oxidation of propane. They also showed that acetone can produce free radicals which sequester reactive nitrogen in the form of PAN. Recent photochemical studies have used one-dimensional and two-dimensional models to describe its atmospheric structure based on its source resulting from the oxidation of molecules such as propane and isobutane [Kasting and Singh, 1986;Chatfield et al., 1987;Singh and Kasting, 1988;Henderson et al., 1989;Kanakidou et al., 1991]. Data to support these model results unfortunately are extremely sparse, in part because of lack of suitable measurement techniques [Vairavamurthy et al., 1992]. Table 1 summarizes available measurements of acetone, most of which were made at rural surface sites using a variety of direct and indirect techniques. The only measurements that are not made on the ground are those of Arnold et al. [1986] who used an indirect chemical ionization technique, applicable in the upper troposphere and the lower stratosphere, to estimate acetone abundances. The wet chemical derivative techniques such as the DNPH/HPLC method have also been used for aldehydes and ketones in semipolluted environments, but these require long sampling times and suffer from potential interferences [Vairavamurthy et al., 1992]. Overall, few reliable measurements of acetone Acetone Mixing Ratio, ppb  [1982] Snider and Dawson [1985] ArnoM et al.

Atmospheric reactions of these NMHCs with ozone (03) and free radicals result in the formation of a variety of intermediate oxygenated species of which carbonyls (RR'C=O) form an important group [National Academy of Sciences
[1986] Shepson et al. [1991] this study ND, not detectable; N, number of data pointsl CIMS, chemical ionizatioh mass spectrometery; DNPH, Dinitrophenyl hydrazine; HPLC, high performance liquid chromatography; RGD, reduction gas detector. are available in the global troposphere and very little about its sources and sinks is known.
During the summer (July to August) of 1990, NASA GTE/Arctic Boundary Layer Expedition (ABLE)3B experiment provided an opportunity to utilize a recently developed instrument to perform real-time airborne measurements of acetone between 35øN and 65øN to an altitude of 6 km. In addition, a comprehensive set of trace chemical (03, NOx, NOy, PAN, HNO 3, C2-C 6 NMHC, CO, halocarbons) and meteorological parameters were also measured. Here we present the first airborne measurements of acetone based on its direct real-time detection and use these to characterize its distributions and variabilities in the troposphere. Relationships between the atmospheric concentrations of acetone and other key chemical parameters are explored. Based on the interpretation of these field measurements and the exercise of photochemical models, we construct a rough picture of its sources and sinks. The PANAK instrument was designed to measure peroxyacetylnitrate (PAN) and select aldehyde and ketones using a pair of airborne gas chromatographs. Results based on the measurements of PAN and other reactive nitrogen species are discussed in a companion paper [Singh et al., this issue]. The initial focus for carbonyl measurements was on acetone and acetaldehyde. The instrument used depended on cryogenic collection of about 300 ml of air, separation of key species with glass capillary column, and sensitive detection by a mercuric oxide reduction gas detector (RGD). The RGD is based on the UV detection of mercury vapor released when a species X reacts with heated (285øC) mercuric oxide (X+HgO(solid) --> XO+Hg(vapor)). It has been previously shown by O'Hara and  that RGD can detect acetone and acetaldehyde (and possibly formaldehyde) with a sensitivity that is 20 to 30 times greater than that of FID. Since the RGD is also sensitive to other species such as alkenes, sulfur compounds, and some higher hydrocarbons, the analytical technique devised used a precolumn (with timeprogrammed backflush) to reduce potential chromatographic interferences.
Species of interest, such as acetone, were cryogenically (-170øC) enriched from a 300 ml (0øC, 1 atm) air sample in a multistage open tube trap connected to a ten-port gas sample valve. The concentrate was then injected into a precolumn containing two stationary phases. The first section of the packed precolumn was 20 cm x 0.18 cm ID section of 10% carbowax 600 on Supelcoport connected to a second section of 30-cm length containing 0.2% carbowax 1500 on Carbopack. The precolumns and valve assembly were held at 42øC. The compounds flushed through the precolumn were cryofocused in second dual stage trap attached to a six-port valve. When acetone was collected in the second trap (time program), the two phase precolumn was backflushed to substantially eliminate potential interferences such as water and high boiling organics. Trapped material from the second precolumn was injected into the main analytical column by heating it to 70øC. The main analytical column itself was a 30 m x 0.75 mm 1/•m Supelcowax 10 glass capillary column fitted with a 0.2 m x 1 mm 10% carbowax 1000 on Supelcoport guard column. A helium carrier gas flow rate of 10 cm3/min and an isothermal oven temperature of 47øC were employed. A make up flow of 10 cm3/min was provided to allow the RGD to receive a total of 20 cm3/min of helium.
In-flight calibrations were performed with the help of an acetone permeation tube (VICI-Metronics) held at 0øC in an icepoint flight dewar. This tube permeated at a mean rate of 58 ng/min of acetone. A dynamic dilution system using compressed ultrazero air was used to generate acetone standards in the 50 to 3000 ppt range bracketing actual tropospheric measurements. To ensure that the permeation tube was uncontaminated and behaved properly, laboratory standards were prepared in a variety of liquid medium and compared with the acetone standards generated from the acetone permeation tube. These comparisons agreed to typically within 10% and confirmed the purity of the acetone emission. It is noted that the same studies also showed that the acetaldehyde permeation tube did not work properly and the gravimetric loss rates did not correspond to acetaldehyde emissions over a period of several weeks. One of the causes was found to be the slow oxidation of acetaldehyde to acetic acid. In the field, all calibrations were performed with the acetone permeation system. As is shown in Figure 1, the detector is linear in the concentration range of interest. ABLE3B provided the first opportunity for the deployment of this GC-RGD system aboard an aircraft. During field operation the system had a sensitivity of 10 ppt acetone (300-m! standard sample) and a measurement was made about every 12 min. The sampling frequency was in part dictated by the fact that the RGD system shared its computer control with the PAN instrument. A precision of +10% and an overall accuracy of about +15% are estimated.

RESULTS AND DISCUSSION
Acetone measurements were made starting with the Goose Bay operation and continued till the end of ABLE3B (missions 13-22). An integrated data file, containing acetone data and the corresponding mixing ratios of a large variety of trace species measured by ABLE3B investigators, was created for the purpose of data analysis. These complementary data were averaged within the acetone sampling window. In the analysis and interpretation that follows, it is important to remember that all measurements are not exactly overlapping in time and space.

Atmospheric Distribution of Acetone
Acetone was found to be an ubiquitous component of the atmosphere and was measured in a concentration range of 357 to 2310 ppt. Substantial variability in the atmospheric abundance of acetone was observed and its vertical structure varied from day to day. This variability was associated with meteorological patterns responsible for the transport of acetone and other trace species from a variety of source regions. Since acetone has significant secondary sources and a relatively short lifetime, differences in chemical production and destruction rates can also cause atmospheric gradients and variability. As is discussed later in this paper, atmospheric lifetime of acetone is longer than 10 days and long-range transport is clearly possible. The mean acetone concentration observed during this entire effort was 1140 + 413 ppt (n=123). Figure 2 shows the frequency distribution of acetone and ethane based on data collected during ABLE3B. Ethane is thought to be the most abundant nonmethane organic species in the global atmosphere [Singh and Zimmerman, 1992]. It is clear from Figure 2, that acetone was more abundant than ethane during this entire experiment. The possibility clearly exists that oxygenated species (such as carbonyls, alcohols, ethers) are important components of the global atmosphere but have not yet been fully investigated. Figure 3 shows the vertical structure of acetone and ethane as measured in the 45ø-65øN latitude belt. The mean acetone vertical profile is nearly constant with substantial variability at all levels. Frequently mixing ratios were higher aloft than near the ground. The lowest mixing ratios were found during clean Arctic air episodes and coincided with low CO, NMHC, C2C14, and 0 3 values. Table 2 shows data involving two such periods encountered in the free troposphere (4-6 km, mission 14) and the boundary layer (0-2 km, mission 18).

Trajectory analysis performed by Shipham et al. [this issue] indicate that the air masses encountered during mission 14
probably originated in the subtropical Pacific some 15 days earlier and were influenced by the Arctic/subarctic air en route to the sampling area. The lowest mixing ratios observed in the free troposphere were in the 500-to 600-ppt range, while the lowest levels encountered in the surface air were in the 350-to 400-ppt range. The latter were found in clean Arctic air masses and over water bodies (e.g., missions 18 and 19 over Frobisher Bay). As will be discussed later in this paper, it is possible that important amounts of acetone are exchanged (but probably not lost) across water bodies. Table  2 also shows the extremely low levels of other primary and secondary tracers present in these air masses. For example, the PAN concentrations measured in this 4-to 6-km altitude were among the lowest ever encountered. It is evident that background mixing ratios of at least 400 to 600 ppt could be found in much of the troposphere studied during ABLE3B (0-6 km). The surface mean of 1140 ppt and clean background mixing ratios of =500 ppt are in general agreement with the limited surface data from the Arctic reported by Cavanagh et al. [1969] and over the Atlantic reported by Penkett [1982]. The results in this study, however, show significantly higher acetone abundances at 6 km than those indirectly estimated by Arnold et al. (CO, NMHCs, NOy, and C2C14). Figure 5 shows the latitudinal behavior of acetone, CO, and C2H2 in the free troposphere for altitudes above 4 kin. The CO and C2H2 mixing ratios are median values from a 5 ø latitude band. It is evident that free tropospheric concentrations of all these species are larger at the subpolar latitudes. In large part this is due to the slower removal rates and the influence of northerly sources such as fires.

Relationships Between Acetone and Selected Tracer and Oxidized Species
As has been previously stated, the measured concentration of acetone showed significant variability above its geochemical background of 400-600 ppt. The relationship between acetone and other primary and oxidized species is explored to get a better understanding of the nature of its sources. Figure 6 shows these correlations with select molecules such as C2H 2, CO, C3H8, and C2C14. The first three are tracers of both urban and biomass-burning sources, while the last is purely indicative of urban/industrial pollution. Propane is also an eventual precursor of acetone. The data are segregated to show mixing ratios in the boundary layer (0-2 kin) and aloft (>2 kin). Figure 6 indicates that a significant association between the variability of acetone and that of key species is present. These relation-ships point to the common sources of CO, C2H 2 and C3H8, and acetone. The relationship with C2C14 is weaker in part because of the difficulty in resolving small changes in its measured concentrations of less than 20 ppt (precision=2 ppt) above a background of about 10 ppt. This background itself is somewhat variable because of hemispherical scale gradients in the mixing ratio of C2C14 due to its moderately short lifetime (--6 months). Other tracers such as CFCs showed worse precision problems and were not used. Indeed, air masses were strongly influenced by biomass burning in this region [Shipham et al., this issue]. Figure 7 shows the same relationships with oxidized species such as NOy, PAN, and 0 3 . Although these data are much more scattered, they still suggest that the sources that impact the distribution of acetone also have a substantial impact on NOy, 03, and PAN.
These relationships in the context of ABLE3B support biomass burning and industrial pollution as two important sources of acetone which are also responsible for much of the observed atmospheric variability.

Case Studies of Biomass Burning and Industrial Pollution
Biomass burning as a source of acetone. Air masses strongly affected by forest fire plumes were sampled during ABLE3B. In virtually all cases an enhancement of acetone was observed. Figures 8 and 9 Figure 10) Aacetone/AC2C14 was found to be in the vicinity of 50-60 as opposed to 3 in a fresh plume (mission 21). As we shall see in the next section, the most important source of acetone is secondary in nature, and Aacetone/AC2C14 would increase with the age of the air mass. The secondary sources, however, may take many days to produce acetone and considerable dilution would have occurred. It is noted however, that these estimated 2-to 5-ppb urban acetone mixing ratios are lower than the 12 (+4) ppb reported by Snider and Dawson [1985] in Arizona but comparable to the 0.2-to 3-ppb levels reported by Grosjean et al. [1990] from several Brazilian cities. Isidorov [1990] reports of studies from West Berlin in 1977 where acetone mixing ratios as high as 120 ppb were measured. In all of these cases, substantially different measurement techniques were employed.

Sources, Sinks and Lifetime of Acetone
While the basic steps that determine the atmospheric chemistry of acetone and its precursors are relatively well defined, its sources are not well understood. The job is made further difficult by the fact that very little atmospheric data are available to construct its budget and the removal rates are both a function of season and latitude as acetone is removed by reaction with free radicals, by photolysis and dry and wet deposition. By using both photochemical models and measurements, we make a first attempt to develop a budget for acetone and compare our estimates of sources and sinks to describe areas of uncertainties.

1990].
Although this study was not primarily designed to measure polluted atmospheres, opportunities presented themselves to sample polluted air masses which indicated evidence of fresh as well as aged contamination. During the return flight (August 15, mission 22), acetone concentrations in the marine boundary layer, impacted by the east coast urban plume, of 2.3 ppb were measured. A fresh urban plume was also sampled in the boundary layer during mission 21 (August 14, 2207 to 2210 UT). In this plume, levels of all urban tracers registered significant increases. An in-plume enhancement of 0.2 ppb acetone coincided with an enhance-

Primary man-made emissions. There is little doubt that direct anthropogenic emissions of acetone are present. Acetone is a commonly used solvent that is manufactured in large quantities. In the United States an attempt has been made to determine its stationary source emissions as part of the NAPAP emissions inventory. Its aggregated emissions from use as a solvent and chemical intermediate in the United
States are estimated to be 0.2-0.25 Tg/yr [Middleton et al., 1990]. Acetone has also been identified as a direct emission from automobile tail pipes [Sigsby et al., 1987]. In a study of exhaust from 46 cars it was found that acetone emissions from the tail pipe were highly variable and constituted 10-70% of the total aldehyde fraction which in turn was 2-3% of the total hydrocarbon emission (J. Sigsby Secondary sources of acetone. As has been suggested by Singh and Hanst [1981], nearly 80% of propane on a carbon basis is oxidized by OH radicals to produce acetone; the remainder 20% producing propionaldehyde. The following mechanism shows the formation of acetone from propane.

CH3)2C(O2)C(OH)H 2 + NO--> CH3COCH 3 + HOCH 2 +NO 2 (CH3)2C=CH2 + 03 --> Criege Ozonides --> CH3COCH3+ O2CH2
In many studies involving ozonolysis, acetone has been identified as a product of C4-C 6 alkene oxidation [NAS, 1976]. Indeed, any molecule with a structure similar to (CH3)2=CR 2 is a potential source of acetone. Of course, as these molecules get more and more complex, competing reactions will cause the yields of acetone to drop. It is estimated that isobutene and other similar alkenes can result in the formation of about 2-3 Tg/yr of acetone.
The above discussion also applies to natural molecules which have structures similar to (CH3)2=CR 2. The most abundant of these is probably myrcene. Myrcene has been identified as an emission from a variety of plants and vegetation and is often present in measurable atmospheric concentrations [Isidorov et al., 1985;$ingh and Zimmerman, 1992]. Based on emission studies of deciduous, evergreen, and mixed forests, J. Greenberg and P. Zimmerman (NCAR, Boulder, Colorado, private communication, 1992) find that myrcene emissions range from 0 to 5% of those of •-pinene. The lowest myrcene emissions appear to be from evergreen forests. Plant species such as Scots pine have been found to produce significantly higher emissions of myrcene. [1979] suggest that 4% of myrcene carbon may be converted to acetone. This yield may be on the low side because many intermediates that may produce acetone can be lost to chamber walls. Assuming a 4% yield from myrcene, an acetone source of about 0.2-0.3 Tg/yr can be calculated.

Assuming a 200 Tg/yr global source of •-pinene, myrcene emissions can be estimated to be about 5 Tg/yr. Chamber studies of Arnts and Gay
A much larger but much more uncertain source of acetone, from the OH oxidation of •-pinene, has been proposed in an unpublished work by Gu et al. (1984). In their laboratory studies they find that 15 (+10)% of •-pinene by mole is converted to acetone. Assuming that about half the •-pinene is removed by reaction with OH, a large source of acetone (6+4 Tg/yr) can be calculated. Unfortunately, no independent corroboration of this result can be found in the published literature. Recent attempts to identify acetone as a product of •-pinene oxidation have been unsuccessful (R. Atkinson, private communication, 1992). Clearly more definitive work is needed to ascertain or rule out this potentially important pathway.
Direct biogenic emissions. Acetone has been identified as a direct biogenic emission from a large number of plants. In one study, 22 plant species characteristic of northern hemisphere forests were studied for organic emissions and acetone was found to be emitted in all cases [Isidorov et al., 1985]. Recent studies in the United States (P. Goldan and R. Falls, NOAA, Boulder, Colorado, private communication, 1992) have also seen these emissions in a variety of rural locations in southeastern United States. It appears that evergreens are the most productive source of acetone and the least productive sources of myrcene. These investigators have also performed limited laboratory studies to evaluate emission fluxes. In laboratory studies with Colorado blue spruce, Alberta spruce, loblolly pine, and aspen, acetone emissions were determined to be 2.10 -4 to 3.10 -5 g acetone(C)/g CO2 (C) uptake. A CO 2 plant uptake of 9 (+2).1016 g(C)/yr (gross primary productivity) is considered to be a reasonable estimate for the globe. Assuming that the lowest emission rate is representative of plants associated with 80% of CO 2 uptake and the highest rate, typical of evergreens, associated with 20% of CO2 uptake, an acetone global source of 9 Tg/yr can be calculated. Assuming the lowest emission rate to be representative of all vegetation, a 4 Tg/yr source is possible. To further explore the existence of biogenic sources of acetone we compared its atmospheric behavior during ABLE3B with that of a well-known biogenic hydrocarbon, namely, isoprene. Figure 11 shows the relationship between measured isoprene and acetone mixing ratios for the overlapping periods during ABLE3B. Since isoprene was rarely measured in the free troposphere, these data are from the boundary layer only (0-2 km). The three outlier data points are shown as circles as they represent conditions where the air sample was strongly influenced by an urban plume (mission 21) or a smoke plume (mission 13). While the available data are limited, Figure 11 is strongly suggestive of the biogenic sources of acetone in the sampled region. It is certainly plausible that direct biogenic emissions of acetone are significant. A great deal more research is needed to understand its biochemistry and to have meaningful quantitative data on its biogenic emissions.

Biomass burning source. The sampled area was impacted by local and distant fires burning over the Hudson Bay lowlands and over Alaska [Shipham et al., this issue].
Acetone has not yet been identified as a direct emission from fires. However, these field data (e.g., Figure 8) leave little doubt that this indeed is the case. Acetone in instances where fire plumes were sampled was clearly elevated along with other species such as CO and C2H 2. By carefully analyzing measurements within fire plumes, a Aacetone/ACO(v/v) = 0.025 (0.02-0.03) could be calculated. This acetone enhancement was probably an upper limit as some formation of acetone from hydrocarbon oxidation may already have occurred within the sampled plumes. The same plumes had a AC2H2/ACO ratio in the vicinity of 0.001-0.002 in near agreement with results from studies over Alaska [Wofsy et al., 1992]. It is further noted that the overall mean correlation of acetone and CO is also linear with a slope of 0.03 ( Figure  6a) suggesting that much of the acetone-CO correlation is probably dictated by the influence of fires in the region. Assuming a CO source from biomass burning of 200 Tg/yr [Lobert et al., 1991], and a Aacetone/ACO of 0.025, an acetone source of 10 (8-12) Tg/yr can be calculated. Clearly more work needs to be done to better quantify this source, but preliminary results from this study suggest that acetone is a significant product of biomass combustion.
Sinks and atmospheric lifetime of acetone. Acetone photolysis and its reactions with OH are probably its major loss pathways. Acetone is capable of producing free radicals and intermediate oxygenates. Among the many products of acetone oxidation are PAN, acetic acid, peroxyacetic acid, methyl glyoxal, and complex peroxides. More detailed chemistry is provided elsewhere [Singh and Hanst, 1981;Henderson et al., 1989;Kanakidou et al., 1991], but the following are some of the salient chemical steps in the atmospheric decomposition of acetone: A lesser amount of removal may also occur through wet and dry deposition. Although acetone is highly water soluble, its partition coefficient strongly favors the gas phase (H = 32 M atm -1 at 25øC) and there appears to be no known mechanism for hydration and significant photolytic loss in aqueous medium [Chatfield et al., 1987;Betterton, 1991]. The possibility of acetone loss by microbial processes after dissolution in seawater cannot be ruled out. Based on present information, it appears that gas phase losses are favored and removal by any process involving dissolution should be small. Seawater algae may also provide a small marine source of acetone [Whelan et al., 1982]. No direct measurements of acetone in seawater are available, but essential equilibrium with the atmospheric abundance can be expected.
We have used a two-dimensional photochemical model to calculate acetone lifetimes based on photolysis and OH removal alone. Absorption cross sections used are those of Emrich and Warneck [1988] and the OH field is consistent with a methyl chloroform atmospheric lifetime of 6.2 years.
Details of the nature of these two-dimensional calculations have already been described by Kanakidou et al. [1991]. Figure 12a shows these calculated atmospheric lifetimes at 55øN. In summer (July 1), acetone lifetime is calculated to be 10-30 days with OH removal dominating in the boundary layer and photolysis dominating in the free troposphere. This lifetime is significantly longer in other seasons approaching 1-3 months in April and 6-36 months in January. Figure 12b shows In (a)the lifetime (for July 1) due to photolysis (hv only)and OH (OH only) is also shown separately .
dimensional and three-dimensional models of the atmosphere were used [Kanakidou et al., 1991[Kanakidou et al., , 1992. In both of these models the only source of acetone was that resulting from propane oxidation. The chemical scheme used in both twodimensional and three-dimensional models described the 03-OH-NOx-CO-CH4-C2H6 and C3H 8 chemistry [Kanakidou et al., 1991]. The model calculations incorporated seasonal and vertical resolution for the OH and the UV field as well as a measure of dry deposition. The deposition velocity (V d) of acetone is probably extremely small and may be negligible [Wesely, 1989]. In the three-dimensional model run, a uniform deposition velocity of 0.05 cm/s, typical of similar species, was applied for acetone. In these model runs, nearly   *A large acetone source (=6 Tg/yr) may be postulated from the oxidation of a-pinene based on the kinetic study of Gu et al. (1984). No independent verification of this result has been possible. +Sink estimate is based on scaling up of model output to fit measurements at 55øN. In the model used, acetone loss is due to photolysis (=64%), OH reactions (=24%), and dry deposition (=12%).

Acetone Global Budget
40-60 Tg/yr. However, this calculation is constrained by rather limited atmospheric data largely collected from the northern hemisphere in one season. Based on estimates shown in Table 3, the source-sink relationship for acetone is roughly in balance but the uncertainties are quite large and in many cases unquantifiable. Clearly a body of atmospheric data that defines the seasonal, vertical, and latitudinal distribution of acetone is needed. Much more accurate information on the direct emissions of acetone from vegetation and biomass combustion are required to further improve the budget of acetone. It is possible that atmospheric abundances of acetone can also be derived from the Jungfraujoch spectra [Ehhalt et al., 1991] and this possibility should be further explored.

Acetone As a Source of Peroxyacetylnitrate (PAN)
Singh and Hanst [1981] have proposed that photolysis of acetone produces peroxyacetyl radicals which can lead to PAN formation. A zero dimensional model as used by Singh and Hanst [1981] and new rate constants as presented by Dernore et al.

CONCLUSIONS
Acetone, an oxygenated organic hydrocarbon, has been found to be the dominant nonmethane organic species present in the subarctic atmosphere sampled during ABLE3B. A substantial global source in the vicinity of 40-60 Tg/yr is estimated to be present. It is suggested that nonmethane hydrocarbon oxidation, biomass burning, and direct biogenic emissions are important sources of acetone. Insufficient data are available to reliably quantify these source strengths. By using atmospheric data and photochemical models, a rough atmospheric budget of acetone has been presented. While the source-sink equation appears to be roughly balanced, much more atmospheric and source data are needed to reliably estimate these relationships. Atmospheric abundance of acetone is expected to have strong latitudinal and seasonal variations, but no direct observations are available. The possibility of deriving acetone abundances from atmospheric spectra (such as the Jungfraujoch spectra) should be explored. It is postulated that acetone photochemistry may be responsible for a significant fraction of the observed free tropospheric PAN. Oxygenated molecules are important components of the atmosphere even though very little about their sources and fate is known.