Characteristics of the atmospheric CO2 signal as observed over the conterminous United States during INTEX-NA

[ ] High resolution in situ measurements of atmospheric CO 2 were made from the NASA DC-8 aircraft during the Intercontinental Chemical Transport Experiment–North America (INTEX-NA) campaign, part of the wider International Consortium for Atmospheric Research on Transport and Transformation (ICARTT). During the summer of 2004, eighteen flights comprising 160 h of measurements were conducted within a region bounded by 27 to 53 (cid:1) N and 36 to 139 (cid:1) W over an altitude range of 0.15 to 12 km. These large-scale surveys provided the opportunity to examine the characteristics of the atmospheric CO 2 signal over sparsely sampled areas of North America and adjacent ocean basins. The observations showed a high degree of variability ( (cid:1) 18%) due to the myriad source and sink processes influencing the air masses intercepted over the INTEX-NA sampling domain. Surface fluxes had strong effects on continental scale concentration gradients. Clear signatures of CO 2 uptake were seen east of the Mississippi River, notably a persistent CO 2 deficit in the lowest 2–3 km. When combining the airborne CO 2 measurements with LANDSAT and MODIS data products, the lowest CO 2 mixing ratios observed during the campaign (337 ppm) were tied to mid-continental agricultural fields planted in corn and soybeans. We used simultaneous measurements of CO, O 3 , C 2 Cl 4 , C 2 H 6 , C 2 H 2 and other unique chemical tracers to differentiate air mass types. Coupling these distinct air mass chemical signatures with transport history permitted identification of convection, stratosphere-troposphere exchange, long-range transport from Eastern Asia, boreal wildfires, and continental outflow as competing processes at multiple scales influencing the observed concentrations. Our results suggest these are important factors contributing to the large-scale distribution in CO 2 mixing ratios thus these observations offer new constraints in the computation of the North American carbon budget.


Introduction
[2] Within the short span of six years since the United Nations IPCC Third Assessment Report, the influence of human activities on climate has become more discernable. The increase in anthropogenic greenhouse gas concentrations, notably CO 2 , is now very likely contributing to the higher globally averaged temperatures recorded since the mid-20th century [IPCC, 2007]. Carbon dioxide has increased in the atmosphere since the pre-industrial period (280 ppm) [IPCC, 2001] and emitted CO 2 has been distributed into the atmosphere, oceans, and terrestrial biosphere which serve as active CO 2 reservoirs. In 2005, the global mean CO 2 mixing ratio reached its highest level in the past 650,000 years [Siegenthaler et al., 2005] at 379 ppm following an annual mean growth rate of 1.9 ppm yr À1 during the last ten years (1995 -2005) [IPCC, 2007]. This high mixing ratio suggests strong net sources in the northern high and midlatitudes due primarily to the combustion of fossil fuels and land use change. The Environmental Protection Agency (EPA) carbon inventory report [EPA, 2006] shows that the total U.S. greenhouse gas emissions were 7074.4 Tg of carbon dioxide equivalents, an increase of 15. 8% from 1990 to 2004. [3] Anthropogenic CO 2 emissions were partly offset by carbon sequestration in forests, trees in urban areas, and croplands at a rate of 11% of total emissions in 2004 [EPA, 2006]. Exchange with the terrestrial biosphere is mostly through uptake by photosynthesis and emission from plant respiration and decomposition of organic matter with seasonal variations. Many carbon cycle studies have indicated that the ocean and terrestrial ecosystems in the Northern Hemisphere are a net sink for atmospheric CO 2 at a rate of about 3 Pg C/year [Bousquet et al., 2000;Fung et al., 2005;Goodale et al., 2001;Myneni et al., 2001;Nemani et al., 2003]. The budget for North America is the best constrained of any continent, with a mean uptake of 1.7 ± 0.5 Pg C/year [Fan et al., 1998], mostly due to re-growth of logged forests and farmlands [Turner et al., 1995], enhanced anthropogenic nitrogen deposition [Holland et al., 1997;Hudson et al., 1994;Schindler and Bayley, 1993;Townsend et al., 1996], and the CO 2 fertilization effect [Kheshgi et al., 1996].
[4] Beyond the fossil fuel contribution to anthropogenic emissions, wildfires also play an important role in the carbon cycle by affecting patterns of carbon accumulation and respiration in soils [French et al., 2003;O'Neill, 2000;O'Neill et al., 2002;Richter et al., 2000], and represent a major global source of gases including carbon containing gases and aerosols emitted directly to the atmosphere particularly during high fire years [Andreae and Merlet, 2001;Crutzen and Andreae, 1990;Liousse et al., 1996;Logan et al., 1981;Seiler and Crutzen, 1980].
[5] During the summer of 2004, the Troposhperic Chemistry Program (TCP) at the National Aeronautics and Space Administration conducted the Intercontinental Chemical Transport Experiment -North America (INTEX-NA) airborne science mission [Singh et al., 2006] to understand the impact of long-range transport of trace gases and aerosols on the changing chemical composition of the troposphere. Instrumentation for measurement of CO 2 was integrated on the NASA DC-8 research aircraft for INTEX-NA to examine its large-scale distribution and to provide insight into the factors influencing its summertime distribution over North America and adjacent oceans.

Study Region and Sampling Method
[6] During the July 1st through August 14th 2004 deployment, high temporal resolution (1 s) in situ CO 2 measurements were recorded aboard the DC-8 providing observations over sparsely sampled areas of North America and the North Atlantic. Eighteen science flights were conducted from three major hubs: Dryden Flight Research Center, CA; Mid-America Airfield, IL; and PEASE International Trade-Port, NH. Airborne surveys were flown over a large area bounded by 36°-139°W longitude and 27°-53°N latitude and included sampling between the 0.15-12 km altitude. Flights were conducted during the day and generally during fair weather.
[7] A modified LI-COR model 6252 non-dispersive infrared gas analyzer was used to determine CO 2 mixing ratios. This dual cell instrument achieves high precision by measuring the differential absorption between sample air and a calibrated reference gas that is traceable to the World Meteorological Organization primary calibration standards. The system was operated at constant pressure (250 torr) and had a precision of ±0.1 ppm (1 sigma) and accuracy of ±0.25 ppm. Experimental procedures are described in detail by Anderson et al. [1996] and Vay et al. [1999Vay et al. [ , 2003. Complementary data used here from numerous instruments has a long history of inclusion in the NASA DC-8 TCP payload, and the techniques were essentially identical to those previously described by Jacob et al. [2003]. Our analysis utilized data for in situ CO, O 3 , C 2 H 2 , C 2 H 6 , Figure 1. Regional distribution of CO 2 during INTEX-NA for the altitude regime of 0 -2 km, 2 -8 km, and 8 -12 km. Data were grouped into 1°latitude by 1°longitude bins and then averaged.  . Distribution of CO 2 as a function of flight date. The CO 2 signal observed in flights over the ocean (shown in gray scale for the 1st, 28th, and 31st July flights) exhibited significantly less variability than those over land, demonstrating the influence of active CO 2 sources and sinks on the measurements. C 2 Cl 4 , CH 2 O, NO 2 , CH 3 OH, CH 3 CN, PAN, and HCN. Mission and meteorological descriptions are given by Singh et al. [2006] and Fuelberg et al. [2007], respectively.

Results and Discussion
3.1. Regional-Scale and Vertical Distributions of CO 2 [8] The spatially averaged CO 2 distribution ( Figure 1) displays substantial variability reflecting not only the biospheric uptake in the northern summer, but also the preponderance of human population and industrial activity in the northern hemisphere. These data, plotted as averages within 1°latitude by 1°longitude bins, coarsely represent three different regimes within the troposphere: 0 -2 km (Planetary Boundary Layer (PBL)), 2 -8 km (mid-troposphere), and 8 -12 km (upper troposphere (UT)). August and September are the time of maximum CO 2 uptake in the northern hemisphere in the high and midlatitude regions [Conway et al., 1994]. These biospheric uptake features are shown distinctively in the PBL bin between the 30 -50°N and east of 90°W over the continent. The lowest CO 2 mixing ratios (337 ppm minimum) were observed near the surface over the mid-continental agricultural-intensive states of Illinois and Nebraska. Higher CO 2 mixing ratios are evident in the free troposphere (FT) as well as an east-west gradient demarcated by the Mississippi River. The trend toward higher CO 2 values with altitude reflects remnants of an earlier seasonal cycle that is out of phase with the lower tropospheric signal by 1-3 months; the average time for transport processes to communicate the surface signal throughout the tropospheric column [Nakazawa et al., 1991]. Moreover, convection, stratospheric intrusions, long-range transport (LRT), and wildfires were factors contributing to the variability observed in the FT and will be discussed in later sections.
[9] Dominance of the terrestrial biospheric uptake during summer can be seen in the longitudinal distribution of atmospheric CO 2 shown in Figure 2. With increasing distance from the east coast, the biospheric signal becomes less evident as it dilutes into the regional background over the Atlantic Ocean. Clear signatures of CO 2 uptake were also seen in the lowest 2 -3 km with a diminished gradient with altitude over land. Higher average CO 2 mixing ratios recorded on flights west of the Mississippi River, likely reflect less active vegetation in the dry western part of the country. From the CO 2 distribution for individual flight days (Figure 3), differences in the CO 2 spatial variability over land versus ocean are apparent. The largest change in CO 2 mixing ratios was observed over the continent (66 ppm) on the July 20th flight. This is in contrast to flights over the Atlantic that exhibited substantially less variability, evident by the compactness of the July 31st data. Statistics for each flight are summarized in Table 1 and reveal median CO 2 mixing ratios that decrease with time (377 to 370 ppm) as sampling progressed deeper into the growing season. Minimum CO 2 values for individual flights often occurred 1 km and >35°N.
[10] Perhaps the best view of the influence of the continent on the CO 2 concentration distributions can be seen in Figure 4 where the data are presented for flights upwind of North America, over, and downwind of the continent. On July 1st, measurements were made over the eastern North Pacific preceding deployment to INTEX-NA intensive sites providing a glimpse into upstream CO 2 concentration fields. Here CO 2 mixing ratio observations varied by 7 ppm versus 66 ppm over the continent, exhibiting $2% variability compared to 18% over land. Within the boundary layer, median values were 378 ppm for the Pacific, 366 ppm for the land, and 369 ppm over the western North Atlantic capturing well the influence of terrestrial ecosystems on the observations. Conspicuous is the seeming absence of a notable anthropogenic perturbation over the continent however, the anthropogenic signal is embedded within the observations, obscured by the seasonal activity of the biosphere. A new CO tracer method has been introduced by Campbell et al. [2007] to disentangle the anthropogenic and biogenic signals within the INTEX-NA data set.
[11] Current knowledge about global CO 2 sources and sinks has been derived principally from the distribution of atmospheric CO 2 concentrations, often using inverse modeling approaches [Gurney et al., 2002]. In principle, it is possible to estimate CO 2 surface fluxes from atmospheric CO 2 concentration, provided that atmospheric transport can be accurately modeled. Presently, regional carbon budgets are reconstructed from measurements conducted at $100 observing stations [Chedin et al., 2003] spread inhomogeneously across the globe thus posing a constraint on the reliability of fluxes estimated at sub-continental scales [Gurney et al., 2004;Kaminski et al., 1999;Rodenbeck et al., 2003]. Using only the surface CO 2 observations for this analysis yields an inferred carbon flux that is highly sensitive to the details of the boundary layer dynamics in the transport model [Gurney et al., 2004]. One possible way to reduce the sensitivity of these inversions to poorly represented boundary layer dynamics is to use CO 2 vertical profiles and/or column measurements in addition to surface observations [Stephens et al., 2007;Yang et al., 2002]. Moreover, the effect of vertical transport on CO 2 concentrations near the surface is a major uncertainty in estimating net regional carbon fluxes therefore free tropospheric measurements are beneficial for evaluating model mixing and the resulting flux estimates [Krakauer et al., 2005] (web ppt file).
[12] During INTEX-NA, vertical soundings executed by the DC-8 permitted characterization of CO 2 variations at relevant scales for resolution of atmospheric models and satellites. Vertical profile data acquired during two of the eighteen science flights (Figure 5a) illustrates differences   observed in the atmospheric column over the continental U.S. versus the North Atlantic Ocean. When comparing signatures over the ocean versus land, CO 2 varied <1% in the atmospheric column probed (e.g., July 31st, 2004 (20040731)) compared with 10% (e.g., 20040814), respectively. In the heavily under-sampled mid-to-upper troposphere, where the AIRS instrument on the Aqua satellite is mainly sensitive to CO 2 , sampling of a highly heterogeneous tracer field is evident. Figure 5b characterizes well CO 2 variations in the vertical dimension for the entire mission data set and shows that the CO 2 signal in the column is dominated by the boundary layer and surface emissions/uptake. In the marine case, the variability of CO 2 below 2 km is, in part, influenced by continental outflow above the Marine Boundary Layer (MBL). In situ observations in the upper troposphere show a higher degree of variability (9 -10 km bin) than in the mid-troposphere over both land and ocean. The altitude profile variability of binaveraged CO 2 data reveals that a precision of $1% is necessary for a space-based sensor with maximum sensitivity in the mid-troposphere to resolve the CO 2 signal. The ocean-land separation potentially lends insight into differences observed in retrievals over the land compared to the oceans in the extra-tropics and may be beneficial for future retrieval algorithm development/refinement and calibration/ validation of evolving remote CO 2 sensors [Washenfelder et al., 2006].

Influences on Spatial Variability
[13] Measurements of atmospheric composition provide powerful constraints to improve understanding of surface fluxes of trace compounds and their subsequent fate in the atmosphere [Palmer et al., 2006]. Given that concentrations of atmospheric constituents observed by aircraft have inevitably been influenced by the interaction between atmospheric transport and the spatial and temporal variation of sources and sinks, we invoke CO 2 along with other simultaneous trace gas measurements and supporting meteorological data in the ensuing sections to elucidate the various anthropogenic, natural, and dynamical influences on the observed variability of CO 2 during INTEX-NA.

Biospheric Influence
[14] Since terrestrial ecosystems are major sources and sinks of carbon, quantifying their role in the continental carbon budget requires an understanding of both fast (hours to days) and longer-term fluxes (years to decades). To quantify and understand the variations in CO 2 behavior between the planetary boundary layer and the free troposphere, and to link the aircraft measured CO 2 variations in the troposphere to terrestrial landscape sources of carbon flows and fluxes for major land cover type, CO 2 measurements were explored with remote-sensing data products and GIS-based methods. Several derived products from the LANDSAT, NOAA AVHRR, and MODIS sensors were used to specify spatiotemporal patterns of land use cover and vegetation characteristics. In particular, MODIS products available from daily (e.g., surface reflectance) to 8-day (e.g., Gross Primary Productivity (GPP)) to monthly (e.g., Leaf Area Index (LAI)) to annual (e.g., Net Primary Productivity (NPP)) temporal resolutions were effectively utilized for characterizing the carbon flows and fluxes over heterogeneous landscapes.
[15] Ideally, aircraft measurements of CO 2 can be separated into portions of flux records representing the land cover types through just averaging the CO 2 values for all cover types involved. However, such separation maybe misrepresentative, as (a) different land cover types are often spatially mixed and heterogeneous in a given area, (b) wind further mixes the CO 2 fluxes to and from adjacent land cover types (e.g., forest versus crop), and (c) land cover areas of the same cover type at different distances from the aircraft in the wind direction may contribute differently to the flux. To elucidate the variability in the observed CO 2 mixing ratios, data were averaged based on the distanceweighted cover fraction approach [Chen et al., 1999]. Using latitude and longitude information coincident with the CO 2 measurements from the aircraft data, we first geo-referenced the land use cover data derived from LANDSAT. We then spatially averaged the CO 2 mixing ratio data along the flight track based on the distance-weighted land cover type fraction for each flight segment according to the aircraft speed and direction. For example, for the July 20th boundary layer run over Illinois (Figure 6a), at a nominal aircraft groundspeed of 180 msec À1 , a sampling distance of about 151 km was covered and corresponded to nearly 5036 pixels in the LANDSAT data. The airborne CO 2 data was partitioned and averaged based on individual flight segments over different land cover types to infer variations. Although there are important mass balance approaches such as the atmospheric boundary layer budget method for accounting for CO 2 variations, we could not attempt the same due to measurement limitations (e.g., vertical velocity of the Atmospheric Boundary Layer (ABL) growth (w + ); surface aerodynamic roughness; atmospheric stability parameters; etc. were unavailable). Nonetheless, we were able to observe some significant trends such as low CO 2 mixing ratios in agricultural regions through the distanceweighted averaging approach.
[16] The result from the above analysis shows that variations in CO 2 mixing ratios at lower altitudes were attributed to terrestrial sources of vegetation cover that is highly heterogeneously occupying the landscape (Figure 6b). Interclass variations in CO 2 mixing ratios between different vegetation types suggested averaged lower CO 2 mixing ratios for corn crop compared to others (the averaged CO 2 concentrations were 338-345 ppm for corn, 348 -354 ppm for soybeans, and 352-360 ppm for pasture and woodlots). This is mainly attributed to the C4-dicarboxylic acid pathway of carbon fixation in corn, having a higher rate of photosynthesis than other C3 plants such as soybeans [Crosbie et al., 1977;Curtis et al., 1969;Ehleringer and Pearcy, 1983;Prueger et al., 2004].
[17] In addition to LANDSAT derived biophysical products, MODIS Normalized Difference Vegetation Index (NDVI) products were utilized in this study. NDVI is an indicator of vegetation activity where values near 0 indicate very sparse vegetation and dense vegetation is indicated by values approaching 1. For the July 20th case study over Illinois, low CO 2 mixing ratios were found to correlate well (R 2 = 0.87) with vegetation density (Figure 6c). Although these results suggest significant biospheric uptake of CO 2 from vegetation, a more thorough analysis is required. For example, Wang et al. [2007] has recently shown that atmospheric CO 2 variations are dominated by coherent regional anomalies of up to several 10s of ppm by surface weather (especially temperature, humidity, and radiation) and advection by synoptic-scale winds. For addressing these issues, models involving atmospheric transport are required to be able to generate fine-resolution transport information to resolve the high frequency airborne CO 2 data sets and to quantify regional scale CO 2 sources and sinks [Wang et al., 2007].

Convective Influence
[18] The upper troposphere (UT) is only indirectly related to surface fluxes via lower troposphere transport, and thus the effect of surface fluxes on concentrations in this region is not entirely obvious [Tiwari et al., 2006]. UT CO 2 is typically dominated by the interplay between fossil fuel emissions, and the seasonal cycle of carbon release and uptake by the land vegetation in the NH. It exhibits a similar seasonality as is observed near the ground, but of a lower magnitude with $1 month time delay compared to the surface fluxes. Convective storms, however, represent an important mechanism for the vertical redistribution of atmospheric CO 2 on much shorter timescales; less than one hour compared to weeks or months [Dickerson et al., 1987]. Deep convection such as occurs in summertime thunderstorms is a highly efficient mechanism for the fast transport of air near the Earth's surface to the UT [Chatfield and Crutzen, 1984;Dickerson et al., 1987;Hauf et al., 1995;Pickering et al., 1988]. Typical convective storms have spatial scales of tens of kilometers and vertical velocities as large as 15 m sec À1 [Dye et al., 2000], making their local influence in the UT extremely strong. Although convection may account for only a small percentage of the mass of lower tropospheric air mixed into the UT, the chemical impact can be significant because Boundary Layer (BL) air has a much different chemical composition than the free troposphere [Mullendore et al., 2005].
[19] Convection occurring during the summer of 2004 was examined to determine its influence on the distribution of CO 2 over the INTEX-NA sampling domain. Several studies show that the Upper Troposphere/Lower Stratosphere (UT/LS) was greatly influenced by convective processes during the mission: 13% by Liang et al.  Fuelberg et al. [2007] and that convective outflow was strongest between 9.5 and 10.5 km [Bertram et al., 2007]. Here, we document the occurrence of a series of convective episodes during INTEX-NA and present trace gas measurements from the penetrations of convective clouds. Since the CO 2 concentration is lower in the BL during the Northern Hemisphere (NH) summer compared to aloft, we use this difference in CO 2 mixing ratio to identify air masses convectively transported from the BL to the UT [Huntrieser et al., 2002]. Figure 7 captures vertical profiles of several trace gases (CO 2 , CO, NO 2 , CH 2 O, C 2 H 6 , and C 2 Cl 4 ) simultaneously measured in convectively influenced air parcels sampled during the July 12th flight over the continental U.S. These vertical profiles were considered together with convective influence forecast plots (http://bocachica. arc.nasa.gov/INTEX/coninf_plots.html) and corresponding DC-8 flight track (Figure 8). The coupling of the convective influence calculations with the tracer data indicates that the BL tracers were carried upward in the core of the updraft to the UT within hours of sampling by the DC-8, experiencing little dilution (Figure 7a); in less than a day (Figure 7b); and in 1 to 2 days (Figure 7c). In the fresh convection case (Figure 7a), CO 2 concentrations <375 ppm were observed in the BL and throughout the UT (e.g., 8, 9.5 and 12 km). Concomitant increases in CO, C 2 Cl 4 , C 2 H 6 in the UT having concentrations similar to those measured in BL air, substan- where CO 2 depleted air sampled at 9 km corresponds with enhanced CH 2 O and C 2 H 6 . Observations over Oklahoma while penetrating a convective cell > 1 day old (Figure 7c) suggest that sufficient time has elapsed for dilution of CO 2 to occur in the UT compared with C 2 H 6 that still conveys evidence of convective transport. From observations of CO 2 Figure 7. Vertical profiles of several trace gases (CO 2 , CO, NO 2 , CH 2 O, C 2 H 6 , and C 2 Cl 4 ) simultaneously measured in convectively influenced air parcels (a) within hours (b) in less than a day (c) in 1 to 2 days during on 12th July. during INTEX-NA, Bertram et al. [2007] calculated that the fraction of BL air in fresh convection was 0.11 ± 0.03.
[20] When pollutants are released in or transported to the UT, the longer lifetimes and higher wind speeds greatly expand their range of influence [Dickerson et al., 1987]. Since CO 2 has an atmospheric lifetime on the order of hours to years, rapid convective transport may inevitably play a crucial role in defining the UT global distributions.  Soja et al., submitted manuscript, 2006). Most of fires burned in the Intermontane Boreal forest (Figure 9) which holds the second largest carbon stock in North American boreal eco-zones [Bourgeau-Chavez et al., 2000;A. L. Soja et al., submitted manuscript, 2006]. Fifteen to twenty percent of the total global carbon emissions from biomass burning [Amiro et al., 2001;Cahoon et al., 1994;Conard et al., 2002;French et al., 2000;Kasischke and Bruhwiler, 2002;Soja et al., 2004] are emitted from these zones as the peatlands and soil organic matter (SOM) they contain represent a vast terrestrial carbon reservoir [Alexeyev and Birdsey, 1998;Apps et al., 1993;Zoltai and Martikainen, 1996] available for release to the atmosphere during fire [Kasischke and Johnstone, 2005;Soja et al., 2004;Turetsky et al., 2004]. Indeed, Turquety et al. [2007] and Pfister et al. [2005] reported that 30 Tg CO including 11 Tg CO from peat were emitted from the 2004 fires.

Biomass Burning Influences
[22] Surface burning products injected into the UT and/or LS [Damoah et al., 2006;Turquety et al., 2007] by pyroconvection were transported eastward and southeastward with the jet stream [Fuelberg et al., 2007] and subsequently sampled by the DC-8. Figure 10 depicts enhanced CO 2 (2.7 ppm) and CO (442 ppbv) in an air mass sampled at 7 km over the western North Atlantic on July 18th. The plume was also augmented with elevated CH 3 CN, PAN, HCN, and H 2 CO which is indicative of combustion contributions from biomass burning. A Modified Combustion Efficiency (MCE = DCO 2 /(DCO 2 + DCO) value of 0.86 was calculated for this plume indicating a smoldering combustion source [Korontzi et al., 2003;Sinha et al., 2003;Ward et al., 1996;Yokelson et al., 1999].  fire emissions that have undergone mixing/dilution with transport.
[23] These results support the recent shift from using the traditional flaming to smoldering ratio of 50:50 to 20:80 or 10:90 for calculating trace gas emissions from boreal SOM and peatland burning [French et al., 2003;Kasischke and Bruhwiler, 2002;Soja et al., 2004]. An increase in the smoldering proportion of this ratio represents an increase in trace gas emissions of CO, CH 4 and NMHC, accounting for the unique fuels stored for decades to centuries in boreal ecosystems. Since biomass burning in this region is predicted to increase under current climate change scenarios and has been corroborated in recent reports [A. L. Soja et al., submitted manuscript, 2006;Stocks et al., 1998], knowledge of the percentage of trace gases emitted to the atmosphere could become increasingly important. These results could play an essential role in the carbon inventory and in simulation transport models, because carbon released from biomass burning is significant, particularly during severe fire seasons like those in 2004.

Stratospheric Influence
[24] Variations in stratospheric CO 2 are mainly a reflection of tropospheric transport on surface concentrations. They contain the signature of the seasonal variation of the lower tropospheric environment and the annual increase although the signal is generally weaker $2 -3 ppm) and somewhat lagged (1 -2 months) compared to the surface signal [Boering et al., 1994;Hall and Prather, 1993;Nakazawa et al., 1991]. Through various atmospheric transport pathways and dynamical processes such as stratospheric intrusions and cut-off lows, subsequent exchange with the free troposphere can occur.
[25] With the INTEX-NA period dominated by lowpressure conditions over northeastern North America, the free troposphere in that region was frequently enriched by stratospheric O 3 [Thompson et al., 2007]. Estimates from several studies of the stratospheric contribution to the upper  , submitted manuscript, 2006;Fuelberg et al., 2007;Thompson et al., 2007]. Examination of the standard deviation of the mean CO 2 profiles illustrated in Figure 5b reveals greater variability in the high altitude data compared to the mid-levels for both ocean and land cases, likely a consequence of this contributing influence.
[26] To investigate the CO 2 variations in stratospherically influenced air masses sampled, we filtered the INTEX-NA data set on O 3 > 200 ppbv and O 3 > 300 ppbv, the latter representing occasional, penetration of the lower stratosphere by the aircraft. This filter is substantiated by correspondingly decreased CO, H 2 O mixing ratios and work by J. E. Dibb et al., submitted manuscript, 2006 using Beryllium-7 and the HNO 3 /O 3 ratio to extract stratospherically impacted air masses from the INTEX-NA data set. Median CO 2 concentrations vary less than 1 ppm from the observed remote marine ''background'' value of 376.2 ppm, differing by 0.5 ppm between the two categories compared to a commensurate change in CO of 22 ppbv and 30 ppmv in H 2 O ( Table 2). The altitude range for these events was 8.2-11.3 km. Data for the Aug. 6th flight segment (Figure 11), acquired when the DC-8 traversed a deep trough that produced a low tropopause over North Carolina, shows characteristic signatures of the tracer relationships observed in stratospherically influenced air masses during the summer of 2004. As the O 3 mixing ratio increased, signifying an increasing stratospheric influence, CO 2 values were positively correlated with O 3 , but anti-correlated with CO. During the NH summer, this relationship is to be expected given the phase delay and different amplitude of the CO 2 seasonality between the ground and UT/LS whereas, the fluxes of CO from the surface do not reverse sign seasonally or diurnally like CO 2 .
[27] Measurements and studies have demonstrated that the stratospheric air influencing the INTEX-NA sampling domain above 7 km originated primarily over the central Pacific at high altitude (J. A. Al-Saadi et al., submitted manuscript, 2006) thus contributing to the observed CO 2 spatial variability in the UT. Transport in this region is important for understanding the distribution of greenhouse gases, predicting climate change, and understanding the chemical and radiative balance of this part of the atmosphere [Ray et al., 1999].

Long Range Transport-Asian Pollution Influence
[28] During INTEX-NA, five distinct Asian pollution plumes were sampled by the DC-8 even though summertime trans-Pacific transport is generally inefficient due to slow large-scale flow compared to springtime, a period of  Figure 11. Time series for CO 2 , CO and O 3 showing the stratospheric influence on measurements recorded in the upper troposphere.
[29] The two strongest cases of sampling Asian continental outflow were on the July 1st and August 2nd flights conducted upwind and downwind of the continental U. S., respectively. The intercepted plumes were enhanced in trace gases associated with fossil fuel combustion (CO, C 2 H 2 ), industrial (C 2 Cl 4 ) and biospheric (CH 3 OH) activity that can be invoked to elucidate the CO 2 signal in these events. Figure 12a shows vertical profiles of these chemical tracers on the July 1st flight when pollution originating in NE China was sampled over the Eastern Pacific. The CO 2 concentrations in this plume are often lower than the median CO 2 profile for this flight (Figure 12a) suggesting uptake by surface vegetation in Asia prior to convective lifting and transport. The plume also contains high levels of methanol which has a sizable biogenic source [MacDonald and Fall, 1993;Schade and Goldstein, 2006;Singh et al., 2000].
[30] Deep convection over NE China and lifting in a typhoon east of Japan initiated LRT transport of Asian emissions prior to interception over the Gulf of Maine on August 2nd. The plume contains some mid-tropospheric (6 -8 km) CO 2 concentrations that are below median values and correspond with CH 3 OH enhancements (Figure 12b). Most however, are more representative of the observed free tropospheric CO 2 background values. The longer transport time (5 -9 days) increases the likelihood of dispersion/ dilution with travel over NA and/or mixing with stratospheric air along isentropic surfaces supported by higher DO 3 /DCO ratios [Liang et al., 2007]. Comparatively fewer enhancements in C 2 Cl 4 (t = 68 days) along with the lesser biogenic signal supports contributions from different source regions than the July 1st case.
[31] Low CO 2 mixing ratios were recorded in all five plumes (Table 3) originating over continental East Asia. Observations of CO 2 depleted air containing elevated CO, C 2 H 2 , C 2 Cl 4 , and CH 3 OH are consistent with mixing of biogenic emissions in the anthropogenic outflow. Estimates range from 7 -11% for the influence from East Asia on the U. S. free troposphere composition during INTEX-NA [J. A. Al-Saadi et al., submitted manuscript, 2006;Fuelberg et al., 2007;Liang et al., 2007]. During the summer of 2004, we find that the East Asian influence is a contributor to the observed variability of CO 2 in the free troposphere and this seasonal export has possible implications for the North American carbon budget. 3.2.6. Continental Outflow to the North Atlantic Region [32] Eastern North America, like other highly industrialized regions of the world, is a major source of atmospheric pollutants on a global basis [Cooper et al., 2001;Li et al., 2005;Milne et al., 2000;Whelpdale et al., 1984]. Although a large fraction of the pollutant mass emitted in such a region returns to the surface by the processes of wet and dry deposition, a significant amount is carried out of the region by the prevailing winds. These North American emissions are transported primarily into and removed from the atmosphere over the North Atlantic Ocean [Stohl and Trickl, 1999;Whelpdale et al., 1984] by midlatitude cyclones propagating eastward and poleward before weakening south of Greenland [Cooper et al., 2001;Dickerson et al., 1995;Merrill and Moody, 1996;Moody et al., 1996], heavily influencing the northeastern United States, Canadian Maritime Provinces, and Newfoundland [Fehsenfeld et al., 1996;Millet et al., 2006].
[33] North American outflow events were extracted from the INTEX-NA data set by examining air mass chemical signatures and backward trajectories. We identified seven distinct cases where air masses enriched in CO, O 3 , SO 2 , C 2 Cl 4 , and C 2 H 2 yet depleted in CO 2 had originated over NA prior to sampling over the North Atlantic. An example is provided in a selected time series from the August 11th flight showing coincident enhancements in CO (140 -180 ppb) and O 3 (60 -97 ppb), while CO 2 366 ppm ( Figure 13). Kinematic 5-day backward trajectories reveals a transport pathway from the Great Lakes to the northeast at  !900 hPa with near-surface contact over much of the distance prior to sampling over the Bay of Fundy.
[34] Six of the seven outflow events were sampled north of 43°N and east of 70°W below 2.2 km, mostly within the BL. BL heights were established by examining coincident potential temperature and high resolution water vapor measurements. CO 2 mixing ratios in all of the outflow cases were 1% to 5% lower than the measured remote marine background value of 376.2 ppm. All of the events were associated with frontal activity which was more frequent than normal during INTEX-NA [Fuelberg et al., 2007].
[35] In summary, both biogenic and anthropogenic signals were captured in North American outflow transported downwind of the continent contributing greatly to the variability of atmospheric CO 2 observed within the BL beyond 43°N. In opposite seasons, when winter air mass advection is $1.8 times that in summer [Whelpdale et al., 1984] and the biosphere less active, export of pollution via this major outflow pathway is likely a significant source of anthropogenic CO 2 to the region.

Summary
[36] High-precision airborne measurements of atmospheric CO 2 mixing ratios were obtained throughout the troposphere over North America and adjacent oceans during the summer of 2004. The INTEX-NA observations substantially enhance the available base of information on the regional-scale distributions of CO 2 in the free troposphere. Results emphasize the observed concentrations were the result of myriad competing processes occurring, at times, external to North America. Our abilities to forecast climate change are reliant on an understanding of the underlying processes such as those revealed by the INTEX-NA data set.