Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 A large number of in situ carbon dioxide (CO2) measurements from 5224 flights were taken by commercial airliners from 2005 to 2010. We analyzed the seasonal cycles in tropospheric CO2 in wide areas of the world over the Eurasian continent, the North Pacific, Southeast Asia, and Oceania. In the Northern Hemisphere, large seasonal changes of CO2 in the upper troposphere are found from spring through summer at northern midlatitudes to high latitudes with significant longitudinal differences; seasonally low CO2 mixing ratios are vertically transported from the surface over the Eurasian continent and then transported eastward to the North Pacific. In the Southern Hemisphere, the CO2 in the upper troposphere increases rapidly from April to June, indicating clearly the interhemispheric transport of high CO2 from the Northern Hemisphere winter. The rapid increase in the upper southern lower latitudes is equivalent to about 0.2 Pg increase in carbon. This interhemispheric transport should be adequately represented in general circulation models for source/sink estimates by inverse methods, because it is comparable to the seasonal or net fluxes estimated for a current inversion area size or a typical subcontinental domain. Estimation for transport of CO2 through the high altitudes will be more important than ever with increasing data from aircraft observations.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Atmospheric carbon dioxide (CO2) has been measured systematically (continuously and/or by weekly flask sampling) at an increasing number of surface stations since 1957 [Keeling et al., 1996], beginning with Mauna Loa, Hawaii and South Pole. This global surface network of measuring stations has provided very precise and long-term measurements of surface CO2, disclosing a relatively organized global distribution of seasonal changes in the CO2 mixing ratio (reflecting the biospheric “breathing” of photosynthesis and respiration), along with an exponentially increasing annual mean (reflecting anthropogenic input) [Conway et al., 1994; World Meteorological Organization (WMO), 2007]. In addition, combined with global carbon cycle models of various complexities, these surface CO2 measurements have shown that about half of the CO2 produced by fossil fuel combustion and cement production has stayed in the atmosphere while the remainder has been absorbed by various natural sinks on land and in the ocean [Tans et al., 1990; Manning and Keeling, 2006].
 Although this network of surface stations is global in nature, it is weighted toward the oceanic areas (coastal regions and islands) in the Northern Hemisphere (NH) and remains sparse over the inland continental areas, in the tropics and the Southern Hemisphere (SH). In addition, the observations are generally conducted near the surface, so any clear observational evidence of the temporal variation in the vertical structure is still relatively limited, resulting in an incomplete view of the three-dimensional temporal variation of atmospheric CO2. Recently, however, several satellite observations have started to map the CO2 distribution on a global scale [Chédin et al., 2003; Buchwitz et al., 2007; Yokota et al., 2009]. But the technology is relatively new and faces significant challenges in improving the measurement precision and the vertical resolution to be of any cogent use at the present time. Research aircraft measurements, on the other hand, can and do provide vertical and horizontal distributions of CO2 with sufficient precision to validate transport models, as well as being useful in providing an increased level of constraint in carbon flux estimates by inverse methods. Recent papers have demonstrated the usefulness of aircraft data to estimate the variations and distributions in sources that are hard to be captured by current surface networks [Crevoisier et al., 2010; Pickett-Heaps et al., 2011]. Another example of which involves the use of observed CO2 vertical gradients in the northern midlatitudes to diagnose the surface flux estimates derived from transport models [Stephens et al., 2007]. Global mass balance requirements have led to estimates of tropical sinks/sources that depend sensitively on the representation of the vertical mixing in the transport model, but there have not been sufficiently enough surface CO2 mixing ratio data in the tropics, nor in the mid to upper atmosphere, to constrain the tropical fluxes and the vertical transport of CO2.
 Several observations of greenhouse gases are ongoing by using aircraft. Extensive research aircraft campaigns could observe very detailed distributions in wide latitudinal bands in a short period of time [Pan et al., 2007; Wofsy et al., 2011]. On the other hand, long-term observations with regular flask samplings have been conducted over North America by NOAA/ESRL Carbon Cycle Greenhouse Gasses group [Crevoisier et al., 2010]. By comparison with aircraft campaign, measurements by using commercial airliners constitute a very powerful tool for providing spatiotemporal variations for long-term over continental domain, such as IAGOS-ERI and CARIBIC projects [Volz-Thomas and The IAGOS Team, 2007; Brenninkmeijer et al., 2007; Schuck et al., 2009]. Among the commercial airliner projects, Japan Airliner (JAL) project [Matsueda and Inoue, 1996; Matsueda et al., 2002] conducted very long observations for greenhouse gases by biweekly flask samplings between Australia and Japan since 1993. Toward achieving a more global view of the three-dimensional variations in CO2, we initiated a collaborative research called the “Comprehensive Observation Network for TRace gases by AIrLiners (CONTRAIL)” project [Machida et al., 2008] with JAL. Five JAL airplanes on regular commercial service measure CO2 continuously during each flight. In addition to the vertical profiles of CO2 during ascent and descent, horizontal measurements are obtained along the flight path. The aircraft measurements cover a substantial geographical region, with a wide longitudinal coverage (0°E–115°W) in midlatitudes to high latitudes in NH (Figure 1). The CONTRAIL observation also extends in the north-south direction, along the various JAL flights between Japan, Australia and Southeast Asia. All locations of the airports in the CONTRAIL observation are summarized in Table 1. This observation provides regional vertical/upper atmospheric CO2 data over extensive areas in the Eurasian continent, tropical region and the Southern Hemisphere where the number of surface stations is limited.
Table 1. List of the Airports Where Aircraft Measurements Were Done
London, United Kingdom
Seoul, Republic of Korea
Incheon, Republic of Korea
Las Vegas, United States
Pusan, Republic of Korea
Los Angeles, United States
Honolulu, United States
Mexico City, Mexico
Kuala Lumpur, Malaysia
Alice Springs, Australia
 In this paper, we show the climatological seasonal distributions of CO2 obtained from the CONTRAIL project for the period of November 2005 to December 2010. In section 2, we describe the CO2 data set used in this study. In section 3, methods for air mass classification and calculating the climatological distributions of CO2 in the upper atmosphere are provided. In section 4, we proceed to describe the seasonal CO2 changes in the upper troposphere. In section 5 we discuss the CO2 transport pathways to Southern Hemisphere based on latitude-pressure cross sections. We conclude the paper with a summary in section 6.
2. Airborne CO2 Data
 The CO2 data used in this study were obtained from the ongoing CONTRAIL program. Since much of the measurement procedure related to this program is given by Machida et al. , only a brief description relevant to the present study is given here. Continuous CO2 Measuring Equipment (CME) is placed on board Boeing 747–400 and 777–200 of Japan Airlines and is programmed to operate automatically on a real-time monitoring configuration with feed from the aircraft's flight navigation data. The flow rate and the absolute pressure of the air sample in the NDIR (LI-COR, LI-840) cell inside the CME are maintained at constant values to avoid signal drift. Two standard gases (CO2 in air) of about 340 ppm and 390 ppm, traceable to the NIES-09 scale [Machida et al., 2011], are introduced into the NDIR for calibration every 10 min during ascent or decent and every 20 or 40 min during level flights. Data from the NDIR is recorded every 10 s (corresponding to about 100 m in altitude) during ascent and descent, and every 1 min (corresponding to about 15 km in horizontal distance) at cruising altitude. Stability and linearity of CME have been tested in the laboratory experiments and the analytical precision is estimated within ±0.2 ppm [Machida et al., 2008]. In addition, the CO2 measurements by the CME are compared with the measurements from the flask samples taken by the Automatic Air Sampling Equipment (ASE) when both observations are conducted during the same flight, and show good agreement within ±0.2 ppm indicating high reliability of the CME CO2 measurements [Matsueda et al., 2008]. Using the CME we have been able to obtain upper atmospheric CO2 distributions over large geographical regions from Japan to Europe, Asia, North America, and Australia (Figure 1). A total number of 2,760,043 CO2 measurements obtained during the 5224 flights by the 5 JAL aircraft from November 2005 to December 2010 have been analyzed for this study. The data are averaged to obtain climatological distributions of CO2 after considering the increasing trend, as described in section 3.2.
3. Method for Analysis
3.1. Air Mass Classification for Analysis
 To analyze the CO2 distributions in the upper atmosphere, various meteorological fields (wind, temperature, geopotential) from the Japan Meteorological Agency Climate Data Assimilation System (JCDAS) reanalysis data [Onogi et al., 2007] have been interpolated in time and space to the aircraft positions. Potential vorticity (PV) is derived from the JCDAS horizontal wind, pressure, temperature fields,
where u,v are horizontal wind, f is the Coriolis parameter a function of latitude, g is the gravitational acceleration, and θ is potential temperature. The tropopause height is determined by using 2 PVU (potential vorticity unit, 1 PVU = 106 m2 s−1 K kg−1) [Sawa et al., 2008]. Any CO2 measurements taken inside the boundary layer (defined by the bulk Richardson number less than 0.25 [Troen and Mahrt, 1986]) during the aircraft's ascent and descent have been removed for the analysis, along with those lying inside the layer up to 10 K difference in the potential temperature from top of the boundary layer, because some of them have been contaminated by polluted air near the airports. However, the rejected data (157,566) accounts for only less than 5.6% of the total data. The equivalent latitude coordinate (ϕq) derived from the use of PV on the potential temperature surfaces are employed in the analysis of the western Pacific cross sections,
where A = A(q, θ, t) is the area in which PV is less than q on a particular isentropic surface with potential temperature θ at time t, and a is the radius of the earth. By using the equivalent latitude, the effects of adiabatic transport are removed, allowing us to understand the CO2 distribution more clearly [Hoor et al., 2004; Hegglin et al., 2006; Sawa et al., 2008].
3.2. Method for Calculating the Climatological Distributions of CO2 in the Upper Atmosphere
 As a first step in the analysis of the climatological seasonal cycle of CO2 in space, we investigate the spatial distribution of the seasonal cycle and increasing trend for the study period of 2005–2010. Figure 2 shows the time series of CO2 observed in the upper troposphere (UT) from 8 km to the tropopause over several latitudinal bands from 40°N to 70°N based on the data between Japan and Europe (Figure 2a), 10°S–40°N (Figure 2b) and 30°S–10°S (Figure 2c) between Japan and Australia. Clear and similar seasonal cycles are found every year. To carry out further analysis of the seasonal cycle and the increasing trend, the monthly mean values for each latitudinal band are analyzed by a curve fitting procedure that includes a linear trend component and two harmonics [Matsueda and Inoue, 1996].
 Seasonal variations of CO2 in the Northern high latitudes (40°–70°N) are about 8 ppm in UT (Figure 2a), which are about a half of those observed at surface stations such as Barrow (71.32°N, 156.51°W) or Shemya (52.72°N, 174.10°E) [Conway et al., 1994; Masarie and Tans, 1995]. CO2 mixing ratios over the Eurasian continent have seasonal minimums in July which are slightly earlier than August or September at these stations. CO2 in UT increases relatively steadily till May, while that at surfaces stations suggests slowdown in increases from January to May after rapid increases from September to January. From midlatitudes toward the equator, the seasonal amplitudes of CO2 rapidly decay (Figure 2b), as shown in previous studies [Nakazawa et al., 1991; Matsueda et al., 2002]. Tropical seasonal cycles in UT are comparable to the surface, while those in northern subtropics tend to show slightly smaller amplitudes and slower increases from December to March in the UT compared with surface observations such as Guam (13.43°N, 144.78°E) or research vessels in the Pacific [Masarie and Tans, 1995]. In the Southern Hemisphere, the seasonality of CO2 shows a complicated variation with double increased peaks around June–July and November–December (Figure 2c) [Nakazawa et al., 1991; Matsueda and Inoue, 1996; Matsueda et al., 2002]. Magnitude of seasonal variations in UT is about 2 ppm which are larger than those observed at the surface stations in SH such as Samoa (14.25°S, 170.56°W) or Cape Grim (40.68°S, 144.69°E) [Conway et al., 1994; Francey et al., 1998].
 The trends of CO2 in UT in these wide latitudinal bands show an average rise of 1.9 ppm/yr characterized by small variability of 1.8–2.0 ppm/yr. The observed trend in the upper troposphere is similar to those observed at the surface stations [WMO, 2007]. Since the overall trend appears to be independent of latitude and time over the study period of 2005–2010, we have transformed our data using the following equation:
 This essentially removes the offset relative to the chosen reference year of 2008, allowing us to construct a mean seasonal cycle of CO2 for the study period. The trend in equation (3) is equal to 1.9 ppm/yr. The objective in this study is to obtain the seasonal differences in climatological distributions of CO2 in the upper atmosphere in wide areas. Although small variability of the increasing trend may affect the distributions deduced in this study, interannual variation in CO2 distribution is an issue in the future.
4. Observed Seasonal Cycles of CO2 in the Upper Troposphere
Figure 3 shows examples of the observed monthly distributions of CO2 in the upper troposphere between 8 km and the tropopause for several months. In January, relatively higher mixing ratios are found in the Northern Hemisphere high-latitude regions (Figure 3a). In April (representing the boreal spring), larger north-south gradients of CO2 are found (Figure 3b). CO2 mixing ratios are larger than 388 ppm in the northern high latitudes with the mixing ratio gradually decreasing to 382 ppm in the southern midlatitudes. Changes in the CO2 mixing ratio from April to July (representing the boreal summer) in the tropics and in SH are relatively small, but we see a relatively clear observational evidence of large decreases in the upper troposphere of the northern high-latitude region (Figure 3c), reflecting the upward propagation of the summer photosynthetic absorption of atmospheric CO2 by the surface vegetation. Figure 3d shows the distribution of CO2 in August. Low mixing ratios are found in the higher-latitude regions in the Northern Hemisphere, and compared with the distribution in July (Figure 3c), there is a noticeable decrease in the CO2 in the upper troposphere over the North Pacific. During the boreal autumn, the CO2 in the upper troposphere in SH shows values in the range of 384.0–384.7 ppm, slightly higher than around 383.5 ppm observed in low latitudes, midlatitudes in NH (Figure 3e).
 The spatial distribution of the peak-to-peak difference in the seasonal amplitude (seasonal maximum minus seasonal minimum) based on the observed monthly CO2 mixing ratio shows significant latitudinal dependency, with larger amplitudes in the northern high latitudes and smaller amplitudes in the tropics and in SH (Figure 3f). This latitudinal dependency is in agreement with the ground-based and previous aircraft measurements [Conway et al., 1994; Nakazawa et al., 1991; Matsueda and Inoue, 1996; Matsueda et al., 2002] and simulations for upper CO2 distributions [Olsen and Randerson, 2004; Niwa et al., 2011]. The longitudinal variation is also shown in the CO2 seasonal amplitude in the northern high-latitudinal zones in Figure 3f. Larger amplitudes of 8∼9 ppm are observed over the central and eastern part of the Eurasian continent, while relatively smaller amplitudes (6∼7 ppm) are found over the western side of the continent and the North Pacific. It is also interesting to make note of a relatively large seasonal change in CO2 of about 7 ppm over the Indian subcontinent, compared to the region over the Pacific in the same latitude zone [Patra et al., 2011]. Larger seasonal changes in CO2 in the upper troposphere over the Eurasian continent can be attributed to the vertical transport of low mixing ratios observed in the summer over the continent.
Figure 4a shows the longitudinal distributions of monthly CO2 mixing ratios between 40°N and 70°N from May to August. The mixing ratio over the continental region (20°E–140°E) begins to decrease slightly in May, with a minimum of 385.5 ppm in the region 60°E–80°E. The CO2 depleted region spreads to 120°E in June, reaching a minimum value of about 380 ppm in July. On the other hand, the mixing ratio over the Northern Pacific (160°E–120°W) begins to decrease after a one month delay, reaching a minimum value of 382 ppm in August. The longitudinal variation in the CO2 mixing ratio is largest in July (∼4 ppm), reducing to around 2 ppm in August. The standard deviation for each month increases from about 1 ppm in May to about 2.0–3.5 ppm in July as the mixing ratio itself decreases. However, while the mixing ratios over the North Pacific are still higher than over the Eurasian continent, there is still a large variability of about 2 ppm in July and August. The observed longitudinal distribution of the seasonal variation in the CO2 mixing ratio and its standard deviation in the upper troposphere during the growing season may reflect the nature of the sporadic propagation of the photosynthetically caused low CO2 from the surface to the upper troposphere during the summer. With the dormancy of the vegetative photosynthetic activity and the weak upflow in the winter (from October to April in Figure 4b), the longitudinal variation in the CO2 mixing ratio is mostly less than 1 ppm, accompanied by a much smaller standard deviation than in the summer. The observed CO2 mixing ratios over the Siberian region are highly variable in the summer, even in high altitudes as shown in Figure 4.
Figure 5 shows an example of a longitudinal CO2 distribution obtained during the flight from Paris, France, to Nagoya, Japan, on 8 July 2007. The CO2 mixing ratios in the upper troposphere between 30°E and 70°E show values of about 380 ppm, likely reflecting the “background” concentration in the upper troposphere during this period. On the other hand, very low CO2 mixing ratios of less than 376 ppm are observed in areas 10°–15°E, 100°–115°E and 125°–135°E over the Eurasian continent. These small-scale variability of large fluctuations detected by our instrument with a sampling interval of 1 min (∼15 km interval) is likely caused by the rapid vertical transport of low-concentration CO2 (due to strong photosynthetic absorption of CO2 by the vegetation in the summer) from the surface to the upper troposphere by small-scale convection or frontal uplifting [Chan et al., 2004]. Same kind of CH4, enhancement in upper troposphere by rapid vertical transport was also observed by aircraft observation over Siberia [Tohjima et al., 1997].
5. CO2 Distributions and Deduced Transport Pathways to the Southern Hemisphere
 Next, we present the north-south meridional cross sections of CO2 observed over the western Pacific regions between 100°E and 160°E, showing a clear observational evidence of the north-south seasonal change in atmospheric CO2 across the hemispheres with height (Figure 6). For each month, CO2 distribution above the planetary boundary layer is shown in equivalent latitude-pressure coordinate, from 850 to 150 hPa. Monthly distribution of atmospheric CO2 is influenced by the seasonal change in the carbon source/sink distribution at the surface and in the atmospheric transport (both vertical and horizontal). The impact of the interhemispheric exchange of CO2 is clearly displayed.
 During the boreal winter-spring, high CO2 is confined mainly to the lower half of the troposphere in the midlatitude to high-latitude NH (Figure 6a). Compared with the higher mixing ratios of 386–387 ppm in NH, CO2 in SH shows lower values of around 382.5–383.5 ppm. Thus, we observe a difference of about 3 ppm across the equator in March. Figure 6b shows CO2 for April. We see a larger north-south CO2 gradient across the equator as the CO2 mixing ratio is now higher than in the previous month in the lower half of the NH troposphere. There is also a clear indication of a southward extension of a high-CO2 region (>388 ppm) in the upper NH tropical troposphere. CO2 mixing ratios of greater than 384–385 ppm are observed in the 400 hPa region over the 30°S to 20°S zone by May (Figure 6c), and spread to regions in SH from 30°S to 40°S by June (Figure 6d). Tropospheric CO2 in NH decreases from July to September (Figures 6f). It is interesting to note that slightly lower CO2 follows behind high CO2 in the southern spring; CO2 <384.5 ppm spreads to SH from NH near the equator in August and to 20°S–30°S in September in the upper troposphere. This lower CO2 could cause a slight drop of about 1 ppm in the upper atmospheric in SH in September (Figure 6f). From the boreal autumn to winter, tropospheric CO2 increases gradually in both Hemispheres. Slightly higher CO2 of 386 ppm is found in the southern tropics (20°S–equator) from lower troposphere to around 300 hPa in December (Figure 6g). The latitudinal variation in CO2 is small, and the mixing ratios in the upper troposphere are similar to those in the northern tropics in November and December.
 We could trace the movement of CO2 on the basis of changes in these distributions because there are no significant sources nor sinks in CO2 in the upper troposphere. From April to June (Figures 6b–6d), the high CO2 is transported from the surface to the upper troposphere in the tropical NH, and it brings the seasonal increases of CO2 in UT (Figure 2b). The increased CO2 in tropical NH spreads to SH through a high-altitude corridor, as the equatorial barrier weakens. This rapid CO2 increase in the upper troposphere from April to June causes a unique seasonal cycle of upper CO2 in SH with a maximum around July, in addition to the austral spring maximum around November (Figure 2c). The CO2 distribution patterns shown in Figures 6b–6d suggest that the first maximum in SH is strongly influenced by the atmospheric transport of higher CO2 from NH to SH through the equatorial upper tropospheric corridor. The phase lag in the occurrence of seasonal maximum in CO2 from NH to SH over the western Pacific from March to September is shown in Figure 6h. The seasonal maximum observed in April at lower troposphere in the equator–60°N zone shifts to the NH upper troposphere in May, and then to the upper troposphere in SH by June. From July to September, the maximum in SH moves to lower altitudes at higher latitudes. Also noted is the shift of the maximum from the upper troposphere in the low-latitude regions in NH to the lower stratosphere in the high latitudes cutting across the tropopause [Sawa et al., 2008]. These phase shifts in the seasonal maximum provide strong evidence of an atmospheric transport of high CO2 into SH and the lower stratosphere via high-altitude corridor in the tropical region. It takes about 5 months for high CO2 to move from the surface region in NH (in April) to the upper troposphere in SH and to the lower NH stratosphere (in September).
 Previous studies have suggested the interhemispheric transport of CO2 in the upper troposphere from the NH to the SH, first with observations [Nakazawa et al., 1991]. However, very limited observations for CO2 are conducted to diagnose vertical or temporal structures of interhemispheric transport. A large number of in situ observations here provide seasonal changes at the monthly time scale, which reveal the distributions from distinct gradients in boreal winter-spring to weakening in summer around the equator, and significant transport of high CO2 above 400 hPa. Miyazaki et al.  suggested that a CO2 gradient in the tropical upper troposphere is created by the uplifting of low-level air during boreal winter and spring, and that this tropical gradient weakens because of the mean divergent flow in the Hadley circulations and cross-equatorial eddy transport. This study with high density in the temporal and spatial distribution of our measurements provides the importance of the interhemispheric CO2 transport without ambiguity the rapid concentration increase from April to June in the SH through the upper troposphere, which has been assessed but not validated well in many atmospheric transport models [Jiang et al., 2008; Miyazaki et al., 2009].
 The frequent aircraft measurements over an extended period of time have revealed unique observational features of the CO2 seasonal variation in the tropical upper troposphere. The low-latitude CO2 in SH (equator–30°S) shows a rapid increase of about 2.5 ppm in the upper troposphere (400–200 hPa) from April to June. As a result, CO2 observed in SH is much higher in the upper troposphere than in the lower troposphere and at the surface by 2–3 ppm during these months [Nakazawa et al., 1991; Conway et al., 1994; Matsueda and Inoue, 1996]. This suggests that the SH upper atmosphere contains more CO2 during this season than could be deduced by the surface mixing ratios.
 Here we estimate the changes of carbon amount in the Southern Hemisphere. Atmosphere is divided into 40 cells to contain same amount of air mass in each cell; 8 latitudinal bands range from 90°N–49°N to 49°S–90°S and 5 levels range from 1000–800 hPa up through 200–0 hPa. Our observation well covers latitudinal bands from high latitudes (<61°N) in NH to low latitudes in SH (>30°S) in the middle to upper troposphere (800–200 hPa). Note that this calculation assumes the longitudinally homogeneous distributions, although flight area is limited in the western Pacific region in SH. The estimation suggests the rapid increase that is equivalent to about 0.2 Pg increase in carbon from April to June in the upper southern lower latitudes (equator–30°S, 400–200 hPa). This seasonal change is comparable to the seasonal fluxes estimated by ecosystem models and inverse methods over a typical sub continental domain (such as Boreal Asia, Boreal North America, Temperate Asia) resulting in seasonal amplitudes on the order of 0.5 PgC [Randerson et al., 1997; Baker et al., 2006] and it is also a same order of the magnitude of net sink for the North America (0.5 PgC yr−1) estimated based on aircraft observations [Crevoisier et al., 2010]. Not taking in the interhemispheric transport through the high-altitude corridor could lead to erroneous estimates in the global carbon budget, because the surface flux estimates are sensitive to the way the atmospheric transport is simulated [Stephens et al., 2007]. Even though the high CO2 transported from the NH to the SH has small impact on the seasonal cycles observed at the surface stations, future inverse systems using aircraft observations and satellite data might be critical to estimate the flux distributions.
 Through analysis of the in situ continuous CO2 measurements obtained from November 2005 to December 2010 by the CONTRAIL program, we have a better picture of the unique and clear seasonal cycle of CO2 in the upper atmosphere. The aircraft data have allowed us to obtain a partial three-dimensional seasonal evolution of the atmospheric CO2 mixing ratio. The above analyses of the distributions of CO2 obtained from the frequent observations by aircraft have delineated how CO2 is transported from the surface to the upper troposphere, from land to ocean, and interhemispherically from NH to SH. Despite the significant advances in our understanding of the global carbon cycle and the expansion of the in situ surface CO2 network, there are many regions in the world that remain sparsely observed, causing large uncertainties in the source/sink estimates of these regions [Gurney et al., 2008].
 The CONTRAIL program is continuing, allowing us to obtain longer time series for a more detailed analysis of interannual variations on the carbon cycle. In addition, the growing data set will serve as a useful reference for model comparisons and satellite validations. Among the ongoing aircraft observations, CONTRAIL project is marked by frequent and wide coverage of CO2 measurements. This can contribute the more representative seasonal cycles of horizontal and vertical distributions and short-term variations of CO2 in summer suggested in this study. In addition, observations between Australia and Japan will be useful to analyze the interannual variations in wide latitudinal bands. We aim at the enhancement our understanding of the carbon exchange process between the biosphere, the ocean, and the atmosphere by getting more global and detailed distributions of CO2 and related trace gases in the atmosphere. The continuation and expansion of airborne measurement programs could advances in atmospheric modeling, including validation with current/future satellite observations.
 We would like to acknowledge many engineers of the Japan Airlines, JAL foundation, and JAMCO Tokyo for supporting our CONTRAIL project. We are thankful to Keiichi Katsumata for his technical support of measurement system and to Noritsugu Tsuda for his assistance with the data processing. We also thank Kaz Higuchi for his comments on the manuscript and the three anonymous reviewers for their useful comments and suggestions. The CONTRAIL project is financially supported by the Research Fund by Global Environmental Research Coordination System of the Ministry of the Environment (MOE) in Japan.