Carbonyl sulfide (OCS): Large-scale distributions over North America during INTEX-NA and relationship to CO 2

[ 1 ] An extensive set of carbonyl sulfide (OCS) observations were made as part of the NASA Intercontinental Chemical Transport Experiment–North America (INTEX-NA) study, flown from 1 July to 14 August 2004 mostly over the eastern United States and Canada. These data show that summertime OCS mixing ratios at low altitude were dominated by surface drawdown and were highly correlated with CO 2 . Although local plumes were observed on some low-altitude flight legs, anthropogenic OCS sources were small compared to this sink. These INTEX-NA observations were in marked contrast to the early springtime 2001 Transport and Chemical Evolution over the Pacific experiment, which sampled Asian outflow dominated by anthropogenic OCS emissions. To test the gridded OCS fluxes used in past models, the INTEX-NA observations were combined with the sulfur transport Eulerian model (STEM) regional atmospheric chemistry model for a top-down assessment of bottom-up OCS surface fluxes for North America. Initial STEM results suggest that the modeled fluxes underestimate the OCS plant sink by more than 200%.


Introduction
[2] Although carbonyl sulfide (OCS) emissions make up only a small fraction of the total sulfur emitted into the atmosphere compared to sulfur dioxide (SO 2 ), its comparatively low solubility and long atmospheric lifetime with respect to tropospheric chemistry and photolysis means that a significant fraction should reach the stratosphere. Once in the stratosphere, OCS is oxidized into sulfuric acid and condensed to form stratospheric aerosol [Crutzen, 1976], which is highly effective in reflecting incoming solar radiation back to space, thus enhancing the Earth's albedo [Charlson et al., 1990]. Nevertheless, there are still a great many uncertainties regarding the role played by OCS in the stratosphere [e.g., Chin and Davis, 1995], including an isotope study suggesting that the contribution of OCS to the stratospheric aerosol layer may be relatively minor [Leung et al., 2002]. However, a recent assessment of stratospheric aerosol concluded that in volcanically quies-cent periods, OCS and SO 2 contribute about equally to the stratospheric sulfur budget [SPARC Scientific Steering Group, 2006].
[3] Sources of OCS include both natural and anthropogenic emissions. OCS is directly emitted from oceans, biomass burning, coal combustion, and industrial emissions, as well as being the product of the oxidation of marine and anthropogenic carbon disulfide (CS 2 ) [Kettle et al., 2002a]. Oxidation of marine dimethyl sulfide (DMS) may also be a significant (about 10%) though uncertain contributor to atmospheric OCS [Barnes et al., 1994;Arsene et al., 2001;Watts, 2000]. Ice core -and firn ice -derived records suggest that human activities account for approximately 25% of modern atmospheric OCS levels [Aydin et al., 2002]. Past and recent fluctuations are closely related to changing global anthropogenic sulfur emissions [Montzka et al., 2004].
[4] OCS is removed by terrestrial vegetation, soils, photolysis, and reactions with OH and O radicals, resulting in a drawdown of OCS concentration in the boundary layer, which is typically correlated with low CO 2 [Khalil and Rasmussen, 1984;Mihalopoulos et al., 1989;Chin and Davis, 1993;Thornton et al., 1996;Andreae and Crutzen, 1997;Watts, 2000]. Recent work by Montzka et al. [2007] demonstrated substantial losses from the midlatitude Northern Hemisphere (NH) continental boundary layer during the growing season, which in turn appears to drive the OCS seasonality throughout the rest of the NH, consistent with a strong terrestrial vegetation sink. Such a sink may limit the atmospheric residence time of OCS to about 2 -3 years and also implies a large shortfall in the magnitude of known OCS sources [Xu et al., 2002;Montzka et al., 2007]. [5] The amplitude of the seasonal cycles of OCS and CO 2 are strongly correlated [Montzka et al., 2007], which is consistent with the observation that the uptake of both gases by vegetation follows a common pathway through the stomata [Goldan et al., 1988;Sandoval-Soto et al., 2005]. However, the OCS molecule tends to be irreversibly taken up by photosynthesizing plants, in contrast to CO 2 , which is also ''exhaled'' during respiration [Sandoval-Soto et al., 2005]. Thus further study of atmospheric OCS may provide insight into the mechanisms that control CO 2 sinks, improve the current systematic error in the simulated seasonal flux cycle of the CO 2 , and help to better understand the overall carbon cycle [Montzka et al., 2007;Gausepohl et al., 2006].

Experiment
[6] During July and August 2004, observations of OCS and other species were made from the NASA DC-8 aircraft during the daytime over North America for the Intercontinental Chemical Transport Experiment -North America (INTEX-NA). A major objective of INTEX-NA was to elucidate the sources, transport, and chemical evolution of air masses on transcontinental/intercontinental scales and their impact on air quality and climate. A particular focus for this study was the quantification and characterization of the inflow and outflow of pollution over North America.
[7] Whole air samples were collected at a frequency of 1 to 5 min and analyzed in the University of California, Irvine (UCI) laboratory utilizing gas chromatography. The sampling, analytical equipment, and procedures were almost identical to those described by Colman et al. [2001] and for Transport and Chemical Evolution over the Pacific (TRACE-P) experiment by Blake et al. [2004]. A short description follows.
[8] Air was collected in evacuated 2 L stainless steel canisters aboard the DC-8 and pressurized with a stainless steel (grease-free) bellows pump. Prior to deployment, the canisters were evacuated and subsequently filled with 20 torr (1 torr = 1.332 mbar) of deionized, degassed water to improve the performance of the analytical system and the stability of compounds in the canisters. The canisters were analyzed typically within 1 week of sample collection. Storage tests have shown that OCS is stable (i.e., statistically there was no difference between the first and last samples) in our canisters for storage times of at least 1 week, which is the period when most loss would be expected to occur.
[9] For analysis, each cryogenically preconcentrated sample was partitioned into five different streams, with each stream sent to one of five column-detector combinations optimized for nonmethane hydrocarbons, halocarbons [Blake et al., 2003;Colman et al., 2001], and sulfur gases [Blake et al., 2004]. The combination that was used to quantify the OCS was a DB-5ms column (60 m; ID, 0.25 mm; film, 0.5 mm) coupled to an HP-5973 mass selective detector quadrupole mass spectrometer. The OCS calibration was performed by reference to a commercial standard (Scott-Marrin), with 5% accuracy, and was the same calibration scale as that employed for the TRACE-P experiment [Blake et al., 2004]. The measurement precision for OCS during INTEX-NA was better than 1%, with a detection limit better than 20 parts per trillion by volume (pptv). OCS was always present above its detection limit.
[10] High-precision in situ measurements of carbon dioxide (CO 2 ) were made on the DC-8 by a modified Li-COR model 6252 infrared gas analyzer having an accuracy and precision of 0.25 ppmv and 0.07 ppmv, respectively [Vay et al., 1999]. [11] For an overview of the Summer 2004 INTEX-NA project, see Singh et al. [2006]. (Data are archived at http:// www-air.larc.nasa.gov.)

Observations
[12] INTEX-NA comprised 17 science flights, during which a total of 2924 whole air samples were collected, covering an area from the eastern Pacific to the western Atlantic United States. The eastern United States was the principal focus ( Figure 1). The University of California, Irvine analysis quantified more than 50 different trace gases, including nonmethane hydrocarbons, halocarbons, and alkyl nitrates. However, this paper focuses on the observations of OCS and its relationship to CO 2 during INTEX-NA, with comparison to the Transport and Chemical Evolution over the Pacific observations and initial modeling results.

Sinks
[13] The INTEX-NA observations, shown as averaged 1°Â 1°grid squares ( Figure 2) reveal that the lowest mixing ratios of OCS were found at the lowest altitudes, with higher mixing ratios aloft over the western United States. A significant correlation between OCS and CO 2 exists, with an R 2 of 0.69 for OCS versus CO 2 samples collected over the continent (Figure 3). Stronger correlations are found for several individual flights, mostly flown over vegetated regions of the eastern United States. For example, flight 12 was flown on 20 July 2004 and comprised six lowaltitude legs, three of which were over the North Atlantic and three over different areas of the eastern United States ( Figure 4) and had the highest R 2 value (0.91) for OCS versus CO 2 ( Figure 3). The low-altitude legs flown over the ocean during flight 12 showed no significant drawdown (or obvious large enhancement) of either OCS or CO 2 ( Figure 5). However, all three legs flown in the continental boundary layer (CBL) showed significant drawdown of both gases ( Figure 5).
[14] Trajectories were calculated from exact flight positions for each leg along the DC-8 flight path using a kinematic model [Fuelberg et al., 1996]. According to these trajectories, the air mass sampled during flight 12 over the ocean had spent at least several days in the marine boundary layer (Figure 6), while the air intercepted in the CBL had spent several days at low altitudes over vegetated continental regions of the United States and Canada (Figure 7). These observations serve to illustrate the contrast between the strong sink of OCS and CO 2 at low altitudes for air with a history of terrestrial influence, compared to air masses with principally oceanic influence.

Comparison With TRACE-P
[15] Because TRACE-P was flown in the Asian 2001 spring season, OCS soil and vegetation sinks were expected to be near seasonal lows [e.g., Kettle et al., 2002b]. For TRACE-P, over the western Pacific, OCS mixing ratios were enhanced by at least 10% in samples collected below 2 km altitude, compared to those collected at 8 -10 km (Figure 8a). Similarly, strong gradients were observed for the anthropogenic tracer gas tetrachloroethene (C 2 Cl 4 ) (Figure 8d and in the papers by Blake et al. [2003Blake et al. [ , 2004) and combustion marker CO (Figure 8b), suggesting that boundary layer levels of OCS may have been strongly influenced by continental anthropogenic sources during TRACE-P [Blake et al., 2004]. Mean TRACE-P levels of OCS (as well as C 2 Cl 4 ) over the central and eastern Pacific, at altitudes below about 4 km, were significantly lower than those over the western Pacific ( Figure 8). This was mainly the result of a diminished influence from continental sources [Blake et al., 2004].
[16] In contrast with the TRACE-P observations, the INTEX-NA OCS and CO 2 data show significant biogenic drawdown near the surface. A mean mixing ratio for OCS below 2 km of 410 pptv accounts for a 10% depletion compared to a mean background value aloft of approximately 450 pptv (Figure 8). The corresponding depletion for CO 2 was about 1.6%.

Sources
[17] Biomass burning emissions from Alaskan wildfires were sampled over the NW Atlantic off the coast of Newfoundland during flight 9, 18 July 2004 ( Figure 1). This plume had been transported rapidly across Canada in about 5 d. The OCS/CO 2 enhancement ratio was approximately 12-20 Â 10 À6 vol/vol (depending on choice of ''background'' value), which is at the upper end of the range reported for SE Asian (mostly biomass burning) emissions [Blake et al., 2004;Thornton et al., 1996], but plotted on Figure 3, the samples would not appear as outliers. The corresponding value for the OCS/CO ratio is about 0.09 (0.07-0.11) Â 10 À6 vol/vol, which is at the lower end of the observations for SE Asia [Blake et al., 2004].
[18] A notable feature of INTEX-NA is the high background concentrations of OCS and CO 2 (>500 pptv OCS) observed at high altitudes (8 -10 km) over the Pacific (flight 3) and Atlantic (flight 15) oceans (Figures 1 and 2). These high-altitude enhancements are associated with meteorological back trajectories that originate from Asia [Liang et al., 2007]. Previous observations closer to the Asian continent show that pollution plumes originating from parts of China associated with high levels of coal burning tend to be particularly enhanced in OCS compared to CO [Blake et al., 2004].
[19] Samples with relatively high CO 2 compared to OCS (seen as outliers in Figure 3) were observed during lowaltitude planetary boundary layer (PBL) runs over agricultural regions of the Midwest (specifically, Nebraska and Oklahoma) during the early flight hours of flights 4 and 7 (Figure 1). These outliers may represent air masses that were heavily influenced by the nighttime boundary layer. Ozone levels were very low (<40 ppb) for the flight 4 leg, and anthropogenic hydrocarbons were elevated for the flight 7 PBL leg. These air masses appear to have encountered a large respiration and/or anthropogenic source of CO 2 but no (or relatively low) emissions of OCS, differentiating them from the typical daytime PBL sampling during INTEX-NA.
[20] For most INTEX-NA samples collected over the eastern United States, any anthropogenic emissions of OCS and CO 2 appear to have been successfully masked by their colocated biospheric sink. For example, Figure 9 highlights a flight where the low-altitude air originated over the industrial Ohio River (OR) valley [Hennigan et al., 2006]. Sharp spikes of elevated CO 2 and SO 2 were observed at 15.9 h, 17.1 h, and 18.9 h for the fast response results for those gases and were accompanied by some short-term enhancement in the relatively low-resolution OCS mixing ratio data (Figure 9). The region bordering the OR region contains a large number of power plants that burn fossil fuels, primarily coal. Because OCS is a source of SO 2 , some of the SO 2 associated with the spike could originate from the oxidation of OCS. However, this SO 2 source is likely insignificant as oxidation of OCS is slow in the boundary layer, and the OR region represents a very large SO 2 source (roughly 25% of United States point source direct emissions [Hennigan et al., 2006]).
[21] The only OCS emission factor for the type of combustion representative of the OR area in the literature to date was that for a coal-fired power plant, reporting an OCS/CO 2 ratio of 2.3 Â 10 À6 (= 0.0049 g OCS kg À1 coal burned at the Cherokee Power Plant in Denver, Colorado) [Khalil and Rasmussen, 1984;Chin and Davis, 1993]. Three industrial plumes observed at low altitudes during INTEX-NA had discernable simultaneous peaks of SO 2 ,  Table 1, corresponding to a mean DOCS/DCO 2 enhancement ratio of approximately 5 -7 Â 10 À6 vol/vol. This is significantly lower than the 9 -24 Â 10 À6 vol/vol range for DOCS/DCO 2 in air masses originating over China and SE Asia during TRACE-P [Blake et al., 2004], indicating that emission plumes from the OR may be cleaner as the result of the burning of low-sulfur coal and more modern coalburning technology in the United States.
[22] As stated in section 3.1, the large-scale levels of OCS and CO 2 sampled over the continent appeared to be strongly influenced by the drawdown of both gases to vegetated continental land surfaces. This sink was associated with a typical ratio for DOCS versus DCO 2 (e.g., from flight 12) of about 5 Â 10 À6 vol/vol (R 2 = 0.91) (Figure 3), so it follows that the industrial plume enhancements with enhancement ratios described above and this drawdown ratio could combine to produce the mean slope of about 5.2 Â 10 À6 (R 2 = 0.69) for the linear best fit for all the INTEX-NA samples collected over the United States (Figure 3). Given this mixture of sources and sinks, we turn to computer modeling of these different OCS components to further interpret these data.

Sulfur Transport Eulerian Model (STEM) Model
[23] The STEM regional chemistry model [Carmichael et al., 2003;Campbell et al., 2007] was used for preliminary simulations of OCS on the INTEX-NA domain, driven by available gridded surface fluxes used in previous models [Kettle et al., 2002a] and fixed boundary conditions. The loss of OCS by oxidation with the atmospheric hydroxyl radical was neglected because the reactive lifetime of more  than a year for July in the Northern Hemisphere is much longer than the transport lifetime in the boundary layer of the model domain of several days. The model domain has a 60 km Â 60 km grid resolution and 21 vertical sigma layers that extend from the surface up to the 100 hPa level.
[24] The OCS surface fluxes have a monthly time resolution for the following flux components: terrestrial vegetation sink, soil sink, ocean source, anthropogenic source, and reactive sources from oxidized anthropogenic and ocean sources of DMS and CS 2 [Kettle et al., 2002a]. All are 1°Â 1°spatial resolution (model domain grid cell is 0.5°), except for ocean OCS flux and ocean OCS flux as oxidized CS 2 , which are both at 5°Â 5° [Kettle et al., 2002a]. The spatial distributions for these fluxes for July are shown in Figure 10. The magnitude of the soil sink, total anthropogenic source, and ocean source are approximately 43%, 77%, and 8% of the plant sink, respectively. The anthropogenic emissions are concentrated in the eastern portion of the domain because of emissions from large point sources, while the plant and soil sinks are distributed more evenly across the domain. The INTEX-NA flight paths are concentrated over the eastern portion of the domain as well.
[25] Two model runs were completed, with model run A driven by the gridded fluxes and model run B run by the same fluxes but with the plant uptake doubled. The plant uptake was doubled in model run B because recent measurements from plant-scale experiments [Sandoval-Soto et al., 2005] and a global air-monitoring network [Montzka et al., 2007] indicate that the true plant sink may be 200% or more than Kettle et al.'s [2002a] estimates of the plant sink. We doubled the plant uptake uniformly in space as an initial investigation of the missing sink. However, future investigations may reveal significant spatial variation in how the sink has been underestimated.
[26] The western boundary condition for the model runs was estimated on the basis of data from the 1 July flight  which took place over the Pacific Ocean. The observed OCS concentration for this flight had a mean of 480 ± 10 pptv, while the mean for all flights was 440 ± 40 pptv. The western model boundary condition is fixed at 480 pptv on the basis of the low variation of the observations over the Pacific. Additional observations and forthcoming results from global transport models should be used for more realistic boundary conditions in future model runs.
[27] Defining the other lateral model domain boundaries is more complicated because of the strong influence by terrestrial sinks. The OCS surface fluxes over Canada shown in Figure 10 suggest that the northern model boundary will be highly varied. These lateral boundaries may best be estimated as space-varying, time-invariant surfaces [Gerbig et al., 2003]. However, for the model runs in this study, the lateral boundaries are fixed at 450 pptv on the basis of the mean INTEX-NA observations to the north and east of the model domain of 450 ± 20 pptv.
[28] The top boundary condition was estimated from high-altitude OCS data. The highest STEM vertical level over the ocean is centered at 13.1 km. The highest altitude for INTEX-NA OCS observations were at 11.9 km. Mean OCS observations above 11 km had a value of 460 ± 10 pptv. A fixed value of 460 pptv is used for the top boundary.
[29] The OCS model error is compared in Figure 11 for model runs A and B. For model run A, the low concentrations of OCS observed in the boundary layer are overestimated by the model. This very large underestimation at the surface was not characteristic of previous model runs when comparing INTEX-NA observations with STEM simulations of CO 2 and CO [Campbell et al., 2007]. Doubling the plant uptake in run B improves the agreement for these low-altitude values; however, considerable overestimation remains. The greatest improvements between run A and run B are made near the surface where the model errors are the largest and the influence of the surface fluxes are dominant. The model performance showed a 15% reduction in RMS errors between model run A and B, with RMS errors of 37.4 pptv and 31.8 pptv, respectively. As an alternative to increasing the plant sink, we could have increased the soil sink or decreased the anthropogenic sources in order to improve the model performance. However, the plant uptake is thought to be underestimated by a factor of 2 or more [Sandoval-Soto et al., 2005;Montzka et al., 2007], while the uncertainties in the anthropogenic source is much smaller ($50%), and the soil sink may actually be overestimated instead of underestimated [Watts, 2000]. The original Kettle et al. [2002a] data suggests that the soil sink was 40% of the plant uptake. Increasing this plant uptake by a factor of 2 or more would make plant uptake the dominant flux for the growing season.
[30] A time series of OCS concentrations compares the observed data to the values obtained from the model runs for the 20 July flight (Figure 12). This flight covers a representative sample of the INTEX-NA flight paths but observes some of the largest drawdown of OCS during the entire experiment. Neither of the model runs captures the depth of the terrestrial sink, even with a doubling of the plant uptake flux. The emissions and model resolutions are too coarse to show the low-altitude, anthropogenic emission spikes. Anthropogenic input from Asia was also shown to influence this flight at high altitudes [Liang et al., 2007], which may account for some of the high-altitude (background) underestimation. Underestimation at high altitudes may also be associated with variability in the background concentrations.

Conclusion and Next Steps
[31] Unlike the net Asian OCS source investigated in TRACE-P during winter/spring [Blake et al., 2004], the data presented here suggests that the dominant terrestrial surface OCS flux in the North American summer is the vegetation sink. This sink works effectively and illustrates the large- scale colocated summer drawdown of OCS and CO 2 and how terrestrial sinks dominate during this season at low altitudes over vegetated regions of North America. Initial STEM model results indicate that the current, proposed magnitude of the OCS surface sink may be underestimated by more than 200%.
[32] Further development of the forward OCS model for analysis of the vegetation sink is planned, adding improved boundary conditions, emissions, and chemistry. Large uncertainties in the global sources of OCS may remain as suggested by recent work [Sandoval-Soto et al., 2005;Montzka et al., 2007]. However, atmospheric analysis of OCS may be more promising for regional applications during the growing season, when the OCS plant uptake is dominant over other OCS surface flux components. Furthermore, a combination of OCS and CO 2 analysis could lead to an improved understanding of the terrestrial biosphere sinks. Figure 10. OCS surface fluxes (pmol OCS m À2 s À1 ) for July over North America from the paper by Kettle et al. [2002aKettle et al. [ , 2002b including the (a) plant sink, (b) soil sink, (c) anthropogenic source, (d) reactive source of anthropogenic CS 2 , (e) ocean source, and (f) reactive source of oceanic dimethyl sulfide (DMS). The oxidized CS 2 ocean source and oxidized DMS anthropogenic source were included in the model but are not plotted here because of their small magnitude. Figure 11. Vertical profile of model error for model run A (driven by gridded fluxes) and model run B (plant uptake increased by a factor of 2) for continental DC-8 flights.