We investigate the upper tropospheric distribution of methane (CH4) at low latitudes based on the analysis of air samples collected from aboard passenger aircraft. The distribution of CH4 exhibits spatial and seasonal differences, such as the pronounced seasonal cycles over tropical Asia and elevated mixing ratios over central Africa. Over Africa, the correlations of methane, ethane, and acetylene with carbon monoxide indicate that these high mixing ratios originate from biomass burning as well as from biogenic sources. Upper tropospheric mixing ratios of CH4were modeled using a chemistry transport model. The simulation captures the large-scale features of the distributions along different flight routes, but discrepancies occur in some regions. Over Africa, where emissions are not well constrained, the model predicts a too steep interhemispheric gradient. During summer, efficient convective vertical transport and enhanced emissions give rise to a large-scale CH4 maximum in the upper troposphere over subtropical Asia. This seasonal (monsoonal) cycle is analyzed with a tagged tracer simulation. The model confirms that in this region convection links upper tropospheric mixing ratios to regional sources on the Indian subcontinent, subtropical East Asia, and Southeast Asia. This type of aircraft data can therefore provide information about surface fluxes.
 Aircraft measurements of upper tropospheric CH4 mixing ratios are carried out sporadically over short periods within larger field projects. Systematic observations up to 7 km altitude are performed at a large number of sites with a focus on the US (http://www.esrl.noaa.gov/gmd/ccgg/aircraft/index.html) and in Siberia [Machida et al., 2001]. Within both programs air samples are collected and analyzed for CH4postflight. Recently the HIAPER Pole-to-Pole Observations (HIPPO) program was concluded after completing a sequence of five long-distance transects over the Pacific 2009–2011 [Wofsy, 2011].
 Measurements from aboard passenger aircraft provide a wealth of information, closing the gap between monitoring by ground station networks and observations by satellites. Especially in the tropics, where strong convection directly links upper tropospheric composition to surface emissions, aircraft data have been used to improve surface flux estimates of greenhouse gases [Patra et al., 2011a; Kort et al., 2011; Baker et al., 2012; Niwa et al., 2012]. CARIBIC (Civil Aircraft for the Regular Investigation of the atmosphere Based on an Instrument Container) [Brenninkmeijer et al., 2007] and CONTRAIL (Comprehensive Observation Network for Trace gases by Airliner) [Machida et al., 2008] are two long-term projects monitoring the upper troposphere and, at midlatitudes, also the lowermost stratosphere on a regular basis. Their operation includes the collection of air samples which are analyzed for a large number of trace gases, among them CH4. Regular data collection over the last six years covering all seasons along different routes over Asia, Africa and the Atlantic Ocean (CARIBIC) and the Pacific (CONTRAIL) provides the basis for a detailed analysis of the upper tropospheric distribution of this greenhouse gas, with a focus on the Northern Hemisphere (sub)tropics, and also in the Southern Hemisphere. Combining these two comprehensive data sets, which need to be analyzed critically in view of temporal and spatial gaps of information, we obtain a broad view of the spatial distribution of CH4 in the upper troposphere. Using a tagged model simulation, the upper tropospheric patterns can be related to surface emissions. In particular, the seasonality in tropical Asia is investigated with the aim to understand the relative contributions of different source regions.
 CARIBIC uses a specially designed air freight container (total weight 1.6 t), equipped with instruments, which is regularly loaded onto a Lufthansa A340-600 passenger aircraft fitted with a multiprobe air inlet system [Brenninkmeijer et al., 2007]. Routine deployment of the current system started in May 2005 and is ongoing. CARIBIC flights take place out of Frankfurt (Germany), typically 2–4 consecutive flights per month. The extensive instrumentation suite includes three units for the automated collection of air samples [Schuck et al., 2009]. Two, which have been in use since 2005, contain 14 glass flasks each; the third, housing 88 stainless steel flasks, was added in 2010, and since then a total of 116 samples is collected per series of flights. Samples are collected at cruise altitude at predefined time intervals. For all air samples several gas chromatography analyses are performed postflight. These include simultaneous gas chromatography measurements of the greenhouse gases CH4, CO2, N2O, and SF6 [Schuck et al., 2009]. CH4 mixing ratios are reported on the NOAA2004 scale [Dlugokencky et al., 2005] with an average analytical precision of 2.2 ppb.
 CONTRAIL employs Automatic Air Sampling Equipment (ASE) onboard two Boeing 747–400 passenger aircraft operated by Japan Airlines. Sample collection is performed about twice per month at predefined locations. During each measurement flight, two sets of six titanium flasks are pressurized with ambient air [Machida et al., 2008]. Laboratory analysis of samples comprises nondispersive infrared measurements of CO2 and gas chromatography measurements of CH4, CO, H2, N2O and SF6. The CH4 measurement precision is 1.7 ppb, and values are reported on the NIES 94 scale, which differs from the NOAA2004 scale by 3.5–4.6 ppb in the 1750–1840 ppb range [Zhou et al., 2009]. A scaling factor of 0.997 was applied to the CONTRAIL data for comparison with the CARIBIC data. This value was derived as part of a scale intercomparison within the GLOBALVIEW-CH4 project (http://www.esrl.noaa.gov/gmd/ccgg/globalview/ch4/ch4_method.html).
 The measurements are compared to results of the Center for Climate System Research/National Institute for Environmental Studies/Frontier Research Center for Global Change Atmospheric General Circulation Model (CCSR/NIES/FRCGC AGCM) based Chemistry Transport Model (ACTM) [Patra et al., 2009, 2011b]. The horizontal model resolution was set to 2.8° × 2.8° (T42 spectral truncation), and 67 sigma pressure levels up to 90 km altitude were used. The meteorological fields were nudged to the 25 year reanalysis data provided by the Japan Meteorological Agency (JRA-25 reanalysis data) over the simulation period of 2005–2012. The spin-up run was repeated 10 times for the year 2004 so that CH4 concentrations reached equilibrium state, both seasonally and spatially. This is reasonable because the lifetime of CH4 in the atmosphere is about eight years due to the loss by chemical reactions with OH, O(1D) and Cl. The surface CH4 flux estimates are equivalent to those in Patra et al. , combining scaled natural/biogenic emissions from the Goddard Institute for Space Studies (GISS) [Fung et al., 1991] and anthropogenic emissions from the Emission Database for Global Atmospheric Research (EDGAR, version 3.2) [Olivier and Berdowski, 2001]. The natural fluxes vary seasonally and are repeated for all years (cyclostationary), but anthropogenic emissions vary in annual time intervals with a small increasing trend with time. The global net annual surface emissions are 520.0, 520.0, 520.0, 521.4, 521.3, 522.4, 522.4 Tg-CH4for the period 2005–2011, after subtracting the soil uptake of 27.2 Tg-CH4/yr. CH4 losses by reaction with OH, Cl and O(1D) are modeled following the TransCom-CH4 protocol [Patra et al., 2011b]. The tropospheric OH field is taken from Spivakovsky et al. , and scaled for successful simulation of CH3CCl3 changes in the atmosphere. CH4 mixing ratios were extracted from hourly model output at the times and locations of air sample collection, and the model data were processed in the same way as the measurements to derive latitudinal distributions and seasonal cycles.
 For a better understanding of regional contributions, a tagged tracer model run was performed in the same manner as in Umezawa et al. . For this tagged simulation, the original surface flux field was divided into 18 regions as illustrated in Figure 1. The flux from each region was simulated separately in the model, and the sum of the 18 individual tracers was consistent with the original global fluxes within 0.1%.
3. Data Overview
 The CARIBIC container is deployed monthly on flights out of Germany to destinations in the Americas, Asia and Africa. On several flight routes the northern midlatitudes and the Northern Hemisphere tropics are probed, and individual flights also go into the Southern Hemisphere. Air sampling takes place at cruise altitude, with over 99% of samples being collected at altitudes between 8 km and 12.5 km. In the tropics, the aircraft flies in the free troposphere, whereas in the extratropics this altitude range corresponds to the tropopause region, and the aircraft frequently encounters stratospheric air masses. Using N2O as an indicator, air samples are classified as tropospheric or stratospheric. Mixing ratios of N2O increase steadily with time, but its variability in the free troposphere is very small. It has no significant sinks in the troposphere, and is removed by photolysis in the stratosphere, which makes it a good tracer of cross-tropopause transport of air [Boering et al., 1996; Hegglin et al., 2006; Ishijima et al., 2010]. To categorize the air samples, the CARIBIC N2O time series is detrended with respect to October 2005, and all samples with mixing ratios below the 2σ range are excluded. This procedure is iterated three times. Only the resulting subset, comprising 2296 samples for which CH4 data is available, is included in the present study. Of these, 825 were collected in the tropics within 30°S and 30°N. These data are complemented by CH4 measurements from 739 tropospheric samples (711 from the tropics), selected in the same manner, which were collected in the upper troposphere at altitudes above 8 km during CONTRAIL flights between Japan and Australia from December 2005 to March 2009. This is only a small subset of the full CONTRAIL data set. CH4 observations on this route have been discussed in detail by Umezawa et al. .
 CARIBIC destinations in the Northern Hemisphere tropics are Guangzhou (23.43°N, CAN, China), Manila (14.52°N, MNL, Philippines), Chennai (12.99°N, MAA, India), Caracas (10.60°N, CCS, Venezuela), and Bogotá (4.70°N, BOG, Colombia). Flights into the Southern Hemisphere went to Santiago de Chile (33.56°S, SCL, Chile) via São Paulo (23.43°S, GRU, Brazil), to Cape Town (33.96°S, CPT, South Africa), and to Johannesburg (26.12°S, JNB, South Africa). The CONTRAIL aircraft conducted flights between Japan and Australia, connecting Narita (35.76°N, NRT, Japan) and Sydney (33.95°S, SYD, Australia) or Brisbane (27.47°S, BNE, Australia). Table 1 summarizes all flights into or crossing tropical regions performed by both measurement programs, and for each route the time period is given. A graphical representation of this information is shown in Figure 2.
Table 1. Months During Which CARIBIC and CONTRAIL Flights Took Place Along Selected Routes Discussed in the Texta
During CARIBIC flights, sampling was performed once per month in both directions. During CONTRAIL flights, samples were collected on the flight leg from Australia to Japan, often twice per month. Destinations are identified by their international three letter airport codes (SA = South America).
GRU 23.43°S/SCL 33.56°S
CAN 23.39°N/MNL 14.52°N
2–8, 10, 11
CCS 10.60°N/BOG 4.70°N
CPT 33.96°S/JNB 26.12°S
SYD 33.95°S/BNE 27.47°S
4. Results and Discussion
4.1. Latitudinal Patterns
 Along the different routes on which the CARIBIC and CONTRAIL instrument packages were deployed, CH4 mixing ratios are found to vary with latitude, in general with lower values in the south. The mixing ratios at cruise altitude are shown in Figure 3 as a function of latitude for flights performed during autumn and winter (October–March). The individual data points have been averaged over latitude bins of 5° width. Figure 3adepicts the distribution along the flight routes to Africa and SA north for winter 2010/2011. Flying between Germany and South Africa, the aircraft stays within a 10°–20° wide longitudinal corridor, and a near-meridional profile is obtained. In the Southern Hemisphere, CH4 mixing ratios are systematically lower. At 30°S, an average mixing ratio of 1787.2 ± 3.7 ppb is measured, while at 30°N it is 1808.9 ± 2.1 ppb. Around and slightly north of the equator, a broad maximum with a peak mixing ratio of 1827.0 ± 4.2 ppb is crossed.
 Middle to upper tropospheric CH4 mixing ratios are also observed by the Atmospheric Infrared Sounder (AIRS) on board the Aqua satellite, data of which can be accessed using the Giovanni online analysis tool on http://disc.sci.gsfc.nasa.gov/giovanni/overview/index.html (accessed February 2012) [Acker and Leptoukh, 2008]. Averaging the observations from November 2010 to March 2011 at 260 hPa, comparable to the CARIBIC flight altitude, relatively low CH4 mixing ratios appear over the Sahara and south of the equator. Over tropical Africa, between 20°N and the equator, mixing ratios are enhanced by about 30 ppb. In the satellite images this maximum is apparent in each November–March period from 2005 to 2012 with an enhancement of 20 ppb to 50 ppb. Satellite observations of the total atmospheric CH4 column densities with the SCIAMACHY instrument also show these higher column densities over the western parts of Central Africa [Schneising et al., 2009; Frankenberg et al., 2011].
 During the same period when the flights to Africa took place, the CARIBIC flight destinations included the northwestern part of South America (SA north), where satellite observations show high CH4 column densities at tropical latitudes [Schneising et al., 2009; Frankenberg et al., 2011]. The aircraft measurements along this route exhibit a very similar latitudinal profile as over Africa, with higher values at northern midlatitudes and an increase in the tropics. At 30°N, the mixing ratios over Africa is lower by 7.7 ppb, and at 40°N it is lower by 5.5 ppb. At all other latitudes the mean values along the flight route to SA south agree within their statistical uncertainties with the measurements over Africa. The increase toward the south occurs near the Venezuelan coast at 12°N, at the same latitude as the maximum over Africa as a consequence of strong continental emissions in the tropics and the enhanced convective activity at these latitudes.
 In Figure 3b the distributions on flights from Japan to Australia and from Germany to Chile via Brazil (SA south), which took place in winter 2005/2006, are compared. In contrast to what was observed over Africa, no CH4 enhancement is encountered when crossing the tropical Atlantic or the western Pacific, while upper tropospheric mixing ratios steadily decrease toward southern latitudes. Between 30°S and 30°N CH4mixing ratios change at a rate of 0.3 ± 0.1 ppb/degree over the Atlantic, whereas over the western Pacific the rate is 0.8 ± 0.1 ppb/degree. Over the complete time period 2005–2009 (not shown) the rate varied between 0.7 ± 0.1 ppb/degree and 0.9 ± 0.1 ppb/degree, and a continuous, though latitudinally variable, year-to-year increase was observed which was strongest from winter 2006/07 to winter 2007/08 [Umezawa et al., 2012].
 Comparing the measurements over South America from 2005/2006 to the more recent ones over Africa in 2010/2011 at 30°S, an increase of 27 ± 9 ppb is measured, at 30°N it is 14 ± 6 ppb. From observations at ground stations annual global average increase rates were derived which, for the full period winter 2005/2006 to winter 2010/2011, add up to 28.4 ± 2.1 ppb (E. Dlugokencky, personal communication, 2012). However, even though the recent growth of CH4 is a well established feature and has been documented over several years by independent observations, the growth rate varies regionally. Based on data from the NOAA sampling network, Dlugokencky et al.  report the strongest growth in 2007 in the Southern Hemisphere and northern polar latitudes, whereas this shifted to the tropics in 2008. Ship measurements in the marine boundary layer over the western Pacific, along a very similar route as the CONTRAIL aircraft tracks, also showed a latitudinally varying growth rate [Terao et al., 2011]. In contrast, the AGAGE and CSIRO observation networks documented a similar rate of increase at all latitudes [Rigby et al., 2008]. The atmospheric transport variations at decadal timescales also introduce a regional growth rate variability between years [Patra et al., 2009]. We conclude that the differences observed around 30°S and 30°N are mainly attributable to the overall growth of methane mixing ratios. At 10°N, in contrast, we measure a mean mixing ratio which is higher by 41 ± 4 ppbv over Africa in 2010/2010 than over the Atlantic in 2005/2006, exceeding the mean global growth.
 The latitudinal distributions along all flight routes were simulated with the aforementioned chemistry transport model. The simulation output was evaluated at the time and location of each individual air sample, after which the mixing ratios were averaged following exactly the same procedure used for the measurement results. The results are thus averaged over the same time period. The resulting modeled latitudinal distributions for the flight routes discussed above are compared in Figure 4 along all routes for boreal winter (Figure 4a) and summer (Figure 4b). The recent increase of CH4 mixing ratios is mainly attributed to increased emissions. Although variations in global OH concentrations may play a role as well, implied by changing mixing ratios of CH3CCl3 [Rigby et al., 2008; Bousquet et al., 2011], global OH is considered well buffered on an interannual timescale [Montzka et al., 2011]. In our ACTM simulation, emissions are kept almost constant after 2005. The global CH4 flux used for 2005 is 520.0 Tg, for 2011 it is 522.4 Tg [Patra et al., 2011b]. In an attempt to account for a potential emission increase, the measurements (solid symbol) are compared to the original model output (solid lines) and to a modified model output (dashed lines). The latter has been derived by adding an average mixing ratio based on the annual growth rates derived from the NOAA sampling network (Dlugokencky, personal communication, 2012). The year 2006 has been used as baseline because the average annual growth rates are given with respect to January 1 and the CARIBIC and CONTRAIL time series start in late 2005. Therefore, no dashed line representing the modified model output is included for flights to SA south in Figure 4a.
 The best agreement between measurements and the unmodified simulation results is obtained for the CONTRAIL flights between Japan and Australia. From an orthogonal regression (not shown) of all individual data points (n = 739) a correlation slope of measured versus modeled mixing ratios of 1.00 ± 0.03 with r2 = 0.55 is obtained, the offset of the intercept differing by less than 1σfrom zero. However, the ACTM-CONTRAIL deviation is larger for higher values of CH4, i.e., for higher latitudes, especially in winter. In winter, the correlation slope is 0.91 ± 0.3 (r2 = 0.51, n = 459), in summer it is 1.14 ± 0.04 (r2 = 0.64, n = 280). In both seasons there is a weak, but statistically significant (99% level) increase with time at all latitudes. This is an indication of increasing CH4 emissions which were regarded almost constant after 2005 in the simulation. For the modified simulation output this trend disappears.
 In Figure 4 it can be seen that the agreement of the original ACTM simulation and the CARIBIC data depends on the flight route and thus indirectly on time. For the route to Brazil/Chile (SA south), which was served by the aircraft in winter 2005/2006, the model and the measurements agree within their statistical uncertainties except for 20°S in the Southern Hemisphere, whereas in the Northern Hemisphere the model overestimates the CH4 mixing ratios by on average 12 ppb. This is similar along the CONTRAIL route to Australia where the model output and the measurement results agree in the Southern Hemisphere but the difference increases from 9.5 ppb at 10°N to 14.5 ppb at 32°N. For the flights over Africa, the model again has a similar, though more pronounced, tendency to overestimate the slope of the interhemispheric gradient, and the discrepancy is largest at the southern part of the latitudinal distribution. Taking into account the global increase of on average 30 ± 2 ppb after 2005 (Dlugokencky, personal communication, 2012), higher values are obtained, and the model results in the Southern Hemisphere agree with the observed mixing ratios, but the gradients are unaltered.
 CARIBIC flights to China, India, and to SA north cover a smaller latitude range, while they took place in summer and winter (Figure 4b). Flying toward SA north, a continuous southward decrease in mixing ratios is observed in summer as well as in winter. When approaching the continent, thus closer to the sources, CH4 mixing ratios increase, and this is more pronounced in boreal winter than in summer. In the latter season, the Intertropical Convergence Zone (ITCZ) is located further north, and the influence from the Southern Hemisphere, with generally lower CH4mixing ratios, reaches beyond the equator into the Northern Hemisphere. This latitudinal pattern is well captured by the model, and the measurement-model deviation does not depend on latitude (no significant correlation). However, the model is systematically low biased, underestimating CH4 mixing ratios by on average 15.2 ppb (16.4 ppb in winter, 14.5 ppb in summer). This difference is larger in later years, related to the underrepresentation of the atmospheric growth rate of CH4. Correcting for this increase in the modified model output, the difference becomes smaller (11.9 ppb), but turns into an overestimation. This implies that in this region the CH4 increase was less than the global average derived from the NOAA ground station network.
 Better agreement is achieved for flights to the Philippines via China (for convenience labeled “China” in Figure 4) which took place mainly in 2006 and 2007 (see Table 1), although the model overestimates CH4 in the two southernmost latitude bins, where over the sea CH4 mixing ratios are lower than over the continent, by 11–29 ppb. Adding the 2007 increase, this becomes even more pronounced. There is little seasonal variation in the shape of the latitudinal distribution along this route. This is different for flights between Germany and India. In summer, a broad maximum of up to 1833.1 ± 7.7 ppb is observed south of 40°N, related to the influence of the South Asian summer monsoon [Schuck et al., 2010], whereas in winter an increase is only seen south of 25°N with a maximum mixing ratio of 1827.6 ± 3.6 ppb. In general, the ACTM largely reproduces the measured CH4 distribution, though in winter it tends to underestimate CH4 mixing ratios by on average 10.5 ppb. In summer, the maximum in the simulation extends to only 30°N whereas the aircraft encounters it up to 40°N. For the modified model output agreement improves for the range 25°N to 35°N, but worsens for 15°N and 40–45°N.
4.2. Measurements Over Central Africa
 While over Asia and over the oceans the model captures the underlying latitudinal gradients along the different flight routes, this gradient is too steep over Africa. Comparing the CH4 mixing ratios at 30°S and 30°N in Figure 4a (first panel) the measured difference is 22 ± 4 ppb, whereas the model yields 37 ± 2 ppb. In addition to CH4, the ACTM simulation also includes SF6 which is a good tracer of atmospheric transport because it is emitted solely from industrial sources and has a very long atmospheric lifetime. While the deviation of the simulation from the measurements increases toward more southern latitudes for CH4, no such latitudinal trend occurs when comparing measured and modeled SF6 (not shown). Considering that most of the sources of SF6 are located in the Northern Hemisphere, this suggests that the model accurately captures interhemispheric transport, and the larger discrepancy for CH4 in the Southern Hemisphere is most likely related to surface flux estimates.
 For a more detailed understanding, a tagged tracer simulation was performed with the ACTM in which emissions were attributed to 18 geographic regions as shown in Figure 1. This revealed that the elevated CH4 mixing ratios measured over Africa are strongly linked to emissions in tropical and southern Africa. Figure 5 shows the seasonality of CH4 fluxes for selected regions, including these two, which have an anticorrelated emission seasonality. During the biomass burning season (NH: October–March, SH: April–July) CH4 emissions are lower than during the rainy season, though emissions north of the equator (tropical Africa in Figure 5) exceed those in the south in all seasons [Roberts et al., 2009].
 Several factors may cause the deviations of the model result from the measurements, surface fluxes, local (small-scale) convection, and large-scale transport. With respect to the latter we gain confidence from the good agreement between modeled and measured SF6 mixing ratios. In order to optimize the shape of the distribution along the CARIBIC flight route over Africa, weighting factors of the individual contributions were estimated in a basic inversion approach, modifying the sensitivity over Africa to the regional fluxes. The latitudinal gradient, which was too steep in the original simulation, largely results from the balance of sources in southern and tropical South America versus temperate North America and Europe. However, the shape of the tropical maximum can only be reproduced by upscaling the contribution from southern Africa by a factor of 1.3–1.6, depending on the choice of parameters. Although this result clearly indicates that CH4 emissions in Southern Africa are underestimated, detailed inversion calculations using surface and aircraft data will be required to quantify the regional emissions.
 The elevated CH4 mixing ratios are also apparent in Figure 6a, which shows the average upper tropospheric distribution over Africa on a 5° × 5° grid, derived from the CARIBIC measurements in winter 2010/2011. Over tropical Africa, CH4 mixing ratios are higher than over the Sahara and over Southern Africa by about 30–40 ppb, similar to the range of 20 ppb to 50 ppb observed by the AIRS satellite instrument. Figure 6b (third panel) repeats the latitudinal profile of CH4between 30°S and 30°N, but here showing values only for December 2010. The highest values are measured around the equator, where also carbon monoxide (CO) mixing ratios, measured in-flight with an ultraviolet fluorescence instrument at a time resolution of 1 s [Scharffe et al., 2012], are elevated. For individual months, the position of the maximum follows the latitudinal shift in upper tropospheric wind patterns, related to the location of the ITCZ and the meandering of the westerly jet. According to air mass back trajectories, the upper tropospheric maximum results from surface emissions in Central Africa north and south of the equator, subsequently subject to rapid convective vertical transport. This is corroborated by the ACTM simulation.
 In addition to CH4, concomitant measurements of ethane (C2H6) and acetylene (C2H2) are available from the air samples [Baker et al., 2010]. These are plotted in the first and second panel of Figure 6b, both showing a similar behavior with latitude as CH4and CO. While the two non-methane hydrocarbons exhibit a tight correlation with CO (r2 = 0.88 for C2H6 and r2 = 0.86 for C2H2), this relation is weaker for CH4 (r2 = 0.64). The slope of the regression line between two trace gases represents a measure of their enhancements relative to background levels, which can be interpreted as an emission ratio [Andreae and Merlet, 2001]. From an orthogonal regression based on 172 samples collected between 30°S and 30°N in the period November 2010 through March 2011, a slope of 7.0 ± 0.2 ppt/ppb is obtained for C2H6 versus CO and 3.1 ± 0.1 ppt/ppb for C2H2 versus CO. Both these values fall within the emission ratios reported for open biomass burning, which range from 4.59 ppt/ppb for grassland burning to 10.77 ppt/ppb for tropical forest burning for C2H6 (4.80 ppt/ppb and 2.17 ppt/ppb for C2H2) [Andreae and Merlet, 2001]. For methane, the slope of the regression line is 0.41 ± 0.02 ppb/ppb, higher than the maximum value of 0.11 ppb/ppb which would be expected for biomass burning (only considering samples collected between 20°S and 20°N we obtain r2 = 0.81 and 0.47 ± 0.03 ppb/ppb for CH4, r2 = 0.87 and 7.8 ± 0.3 ppt/ppb for C2H6, and r2 = 0.84 and 3.5 ± 0.1 ppt/ppb for C2H2).
 While the elevated levels of C2H6 and C2H2 can thus be explained as to originate from biomass burning, the above ratios indicate that part of the observed CH4 is emitted from other sources which do not emit C2H6 and C2H2. CH4 and C2H6 are both emitted from combustion sources, but only CH4 is a product of biogenic sources, such as wetlands which are known to be a strong source of methane in Africa [Bloom et al., 2010]. In boreal winter, when the CARIBIC flights took place, fire activity is highest north of the equator [Roberts et al., 2009]. In contrast, wetland emissions are likely to be dominant in southern Africa, where the ITCZ is located at this time.
 Looking at the individual flight months, the regression slope of CH4 versus CO decreases from December to March from 0.58 ± 0.06 ppb/ppb to 0.38 ± 0.03 ppb/ppb (for November, a smaller number of samples is available and not all correlations are statistically significant). Conversely, the regression slope for C2H6 versus CO increases. Correlating CH4 and C2H6, the regression slope decreases over these months from 0.10 ± 0.01 ppb/ppt to 0.05 ± 0.01 ppb/ppb. In a previous study, these relationships between CH4, C2H6 and CO were applied to CARIBIC data from the flights to India to derive emission estimates for total and biogenic CH4 from the Asian monsoon region, where more CH4 relative to C2H6 was related to biogenic sources [Baker et al., 2012]. Relating mixing ratios of CH4 and C2H6 over Africa yields a value of 103 ± 9 ppb/ppb in December and 87 ± 3 ppb/ppb, similar to the value of 91 ± 9 ppb/ppb that was measured over India at the beginning of the monsoon season. In March, with a mixing ratio of 47 ± 4 ppb/ppb, less CH4 is measured relative to C2H6. This shows that the relative importance of biogenic sources is highest in December and decreases through March, when the rainy season in southern Africa ends, but still the high CH4 to CO ratio of 0.38 ± 0.03 ppb/ppb compared to literature values for biomass burning implies a contribution from biogenic sources.
4.3. Seasonal Variation Over Asia
 The comparison of latitudinal distributions in summer and winter above showed that the seasonality of upper tropospheric CH4 mixing ratios is regionally different, although the distinction of only two seasons does not reveal a complete picture of the seasonal cycle. Boxes in Figure 2 mark some regions of interest in tropical Asia: India south of the Himalayas (10–35°N, 50–80°E), southeast China east of the Tibetan Plateau (20–35°N, 100–115°E), and the South China Sea (15–25°N, 115–125°E), the latter comprising flights between Guangzhou (China) and Manila (Philippines), flying almost exclusively over the sea. For these three regions, CARIBIC CH4 data cover all seasons well so that the annual cycle can be derived. The choice of regional boundaries, which are also listed in Table 2, is largely motivated by the CARIBIC flight destinations and flight tracks to distinguish different meteorological regimes. For the CONTRAIL flights between Japan and Australia a fourth region over the western Pacific is defined (25–35°N, 135–155°E).
Table 2. Regions of Interest Discussed in the Texta
Latitude Range (°N)
Longitude Range (°E)
Region boundaries are defined based on the location of flight destinations, flight tracks, and geographical features.
South China Sea
Figure 7 depicts the seasonal cycle of tropospheric CH4 mixing ratios for these four regions, which are all subject to a monsoonal influence. India in summer is characterized by a southwesterly monsoon near the surface (South Asian or Indian summer monsoon) [Krishnamurti and Bhalme, 1976], carrying clean and humid air with low concentrations of CH4 from the southern Indian Ocean toward the subcontinent [Bhattacharya et al., 2009], while over China and the Philippines the southeasterlies from the Pacific dominate (East Asian monsoon) [Hsu et al., 1999]. At flight altitude, circulation patterns change correspondingly, the most dominant feature being the South Asian upper tropospheric monsoon anticyclone [Hsu et al., 1999; Krishnamurti et al., 2008]. Because of the strong convective activity associated with the summer monsoon, upper tropospheric trace gas mixing ratios are strongly influenced by local surface emissions during summer [Randel and Park, 2006; Xiong et al., 2009; Schuck et al., 2010; Lawrence and Lelieveld, 2010; Baker et al., 2011]. Over the western Pacific, the upper troposphere is under the influence of westerlies, transporting air masses influenced by continental Asian emissions during winter, while this flow is redirected during the summer monsoon period, and slow moving air masses from the central Pacific arrive at the CONTRAIL sampling locations between 15°N and 35°N.
 Over India, CH4 mixing ratios increase throughout summer, peaking in August at a value of 1855 ± 10 ppb. A second, though smaller maximum of 1829 ± 3 ppb occurs in December, following a minimum of 1816 ± 7 ppb in October. Also over southeast China a strong summer maximum of 1880 ± 2 ppb is observed, the peak value occurring in September, while mixing ratios in winter are below 1800 ppb. Over the South China Sea, mixing ratios decrease during early summer, and also reach a maximum in September, though with a mixing ratio of 1807 ± 27 ppb this is lower than the maximum over the continent. While over the sea there are no significant local sources, the air masses probed appear to be influenced by emissions from continental East Asia. These summer maxima over India, China and the South China Sea are a consequence of enhanced emissions during the rainy season in the respective continental source regions, e.g., from wetlands, rice paddies and flooded landfills, combined with enhanced vertical transport due to strong convection [Baker et al., 2012; Umezawa et al., 2012]. This leads to high levels of CH4 in the upper troposphere despite annual maximum concentrations of OH. The timing of the maximum depends on convective activity and also on the seasonality of emissions, both of which are related to the progression of the monsoon and associated precipitation patterns.
 In contrast, farther east, over the western Pacific, mixing ratios start to increase in May and a maximum value of 1820 ± 9 ppb is reached already in July. In August and September, minimum mixing ratios of 1797 ± 6 ppb and 1791 ± 6 ppb occur, contrary to what is measured over India, where upper tropospheric CH4 mixing ratios peak at a value of 1855 ± 10 ppb, and over southeast China with a maximum mixing ratio of 1880 ± 2 ppb in September. This pattern is also linked to the progression of the Asian summer monsoon. In spring, the upper tropospheric monsoon high develops over Southeast Asia from where it moves northwestward [Krishnamurti and Bhalme, 1976; Hsu et al., 1999]. The spring increase is explained by strong convection and enhanced emissions of CH4 from wetlands and rice paddies at the onset of the rainy season in Southeast Asia, when the main outflow from the region is still predominantly directed eastward. In August and September trajectories largely originate over the Central Pacific, which is devoid of CH4 sources. At this point in time the monsoon anticyclone has moved further north, with mainly westward air mass transport toward Africa and the Mediterranean [Lelieveld et al., 2002]. The minimum over the western Pacific can thus be interpreted as a consequence of decreased long-range transport of CH4 from continental Asia, limited by the monsoon anticyclone, southern hemispheric influence close to the equator, and also high summer concentrations of the hydroxyl radical.
Figure 8 shows the measured seasonal cycle for each region in the respective top panels (solid lines) compared to the ACTM result (dashed lines). In addition to Figure 7 with average monthly mixing ratios, here we show monthly deviations from the annual mean. Over India (Figure 8a) and over southeast China (Figure 8b), the model reproduces the seasonality and the timing of the maxima well, although the summer peak values are underestimated by 14 ppb over India and by 31 ppb over China. Over the South China Sea (Figure 8c), the model overestimates the peak value by 33 ppb and the summer maximum occurs in September instead of August. Simulated winter values are 42 ppb too low. Finally, over the western Pacific (Figure 8d), the model captures the general shape of the double peaked structure, but in spring (February–April) values are overestimated by 9–14 ppb, while in summer (June–August) values are underestimated by 8–19 ppb. The first maximum and minimum occur to early in the modeled seasonal cycle by two months and one month, respectively.
 The individual source region contributions, based on the tagging of emissions, are shown for each measurement region in the respective bottom panels. For clarity only those source region contributions that add significantly to the seasonal cycle are shown in color, and contributions without a marked seasonality (less than 5 ppb difference between annual minimum and maximum) are shaded gray. Figure 5 shows the seasonality of CH4 emissions for the most important regions, the boundaries of which are highlighted in black in Figure 1. Over India (Figure 8a), the dominant feature of the seasonal cycle is the summer maximum from July through September. At the same time emissions in India and Bangladesh reach their maximum values which are 60% and 80% higher compared to the fluxes in May, before the onset of the summer monsoon. In previous CARIBIC studies, it was inferred from analysis of back trajectories and tracer-tracer correlations that this is driven by increased emissions from biogenic sources in South Asia and enhanced convective transport in the regions during the rainy summer monsoon season [Schuck et al., 2010; Baker et al., 2011, 2012]. While emissions smoothly decrease from October onward, CH4 mixing ratios decline rapidly in October when the monsoon circulation weakens and increase again to reach the second maximum of 1829 ± 3 ppb in December. During this time of year, the dry monsoon season, continental emissions from Southeast and South Asia, predominantly from biomass burning, biofuel use and the rapidly growing traffic and industry, are transported southward in the boundary layer and transported to the free troposphere in the equatorial convective belt. This pollution transport pathway during the winter monsoon and the effects on the atmospheric composition have been studied in the Indian Ocean Experiment (INDOEX) which took place January–March 1999 [de Laat et al., 2001; Lelieveld et al., 2001; Phadnis et al., 2002].
 The emissions in India and Bangladesh (Figure 5) have a pronounced seasonality with maximum values in late summer. However, for high values to occur in the upper troposphere convective transport has to be strong as well, which is coupled to the evolution of the upper tropospheric monsoon anticyclone. In consequence, the upper tropospheric maximum develops one month earlier than the emissions peak and decreases in October, although emissions are only 5% lower than in September. Interestingly, a large contribution from tropical Africa north of the equator occurs in November. Over India, CARIBIC flights took place in winter 2008/2009 and in winter 2011/2012 (cf. Table 1). The model predicts an increased contribution from tropical Africa compared to summer for both years, however, in November 2011 this was particularly strong. Indeed, ECMWF-based 5 day back trajectories for these flights originate over central Africa, partly indicating boundary layer contact less than five days prior to the measurements. Later in winter the main contribution to CH4 over India is from Southeast Asia.
 The CARIBIC data from flights to China have been used in a cluster analysis based on several meteorological, trace gas, and aerosol particle parameters. This study showed that on this route air masses encountered east of 100°E in summer, i.e., the regions of southeast China and the South China Sea, are strongly influenced by boundary layer air and by high clouds, both an indication of convection. Only a small fraction of less than 5% was found to be undisturbed free tropospheric air, whereas in winter this was the predominant air mass type encountered [Köppe et al., 2009]. According to the ACTM simulation, over southeast China (Figure 8b) comparable contributions from India (14 ppb), China (16 ppb), and Bangladesh (19 ppb) account for the pronounced September maximum, although emissions in China peak already in June, whereas in South Asia the maximum is reached August/September. In October and November emissions from Southeast Asia contribute strongly. Over the South China Sea a more intricate pattern emerges (Figure 8c). The ACTM predicts a similar pattern of the contributions from China, India and Bangladesh, with maximum contributions in June/July and September through November. An overestimation of the latter two contributions, or an inaccurate partitioning between the two regions, may be the cause of the predicted maximum end of October which exceeds the measurements by 33 ppb. In addition, emissions from the northern part of Southeast Asia are predicted to strongly contribute in October, and the model also predicts a considerable contribution from Europe in November. It is conceivable that these two contributions are overrated. In April, emissions from northern Southeast Asia peak, though the contribution from Southeast Asia south of the equator is also significant.
 The seasonality of CH4 mixing ratios over the western Pacific (Figure 8d) is largely explained by emissions from China, exhibiting a clear summer maximum from May through August, coincident with the peak emissions in China depicted in Figure 5. In spring, stronger contributions come from Southeast Asia, and in autumn from India and Bangladesh. From the seasonality of the emissions it would be expected that India and Bangladesh play a more important role already in July, however, as discussed above, these emissions are transported westward by the upper tropospheric anticyclone which is strongest in July and August. After the anticyclone starts its southward retreat in late summer, Indian emissions again exert a strong influence on this region. All four regions are strongly influenced by emissions from China and South Asia, though the transport patterns imposed by the Asian monsoon meteorology strongly modulate the seasonality of upper tropospheric CH4 mixing ratios.
 The regular collection of air samples in the upper troposphere during CARIBIC and CONTRAIL passenger aircraft flights from Germany and Japan, respectively, reveals the distinctive spatial and temporal distribution of CH4 over a large portion of the globe. The present study focuses on data from flights into tropical regions including those into the Southern Hemisphere. Especially in the tropics, upper tropospheric mixing ratios are directly influenced by surface fluxes due to rapid vertical transport of air by convection. Latitudinal distributions are compared in different longitudinal bands, which correspond to the different aircraft flight routes. It is found that upper tropospheric mixing ratios of CH4 show a high degree of variability, despite the CH4 lifetime of nearly a decade.
 A prominent feature are high mixing ratios over Africa in the period October–March. The maximum with a mixing ratio increase of around 40 ppb is centered slightly north of the equator around 0–5°N. While the ITCZ in boreal winter is located in the Southern Hemisphere, it is the biomass burning season north of the equator. Correlations of ethane and acetylene with CO, which all show enhanced mixing ratios parallel to the elevated CH4 mixing ratios, point to biomass burning as an important source. The weaker correlation of CH4 with CO indicates an additional biogenic source, likely wetlands emissions from Africa south of the equator, where for example wetland emissions increase during the rainy season, and, according to the tagged tracer simulation, CH4 emissions are underrated.
 The ACTM simulation results are in overall agreement with the aircraft measurements, although differences appear when comparing in detail. For example, the model overestimates the latitudinal gradient over Africa, and values over southern Africa are up to 27 ppb too low. Interannual variations in the contribution from tropical wetlands [Bousquet et al., 2006, 2011] and also variability in biomass burning emissions, which are typically not very well captured by emission data, may explain some of the differences. Nevertheless, in general the model tends to underestimate CH4mixing ratios in the upper troposphere after 2008. Taking into account that this measurement-model difference increases with time, part of the discrepancy is related to changing surface fluxes after the period covered by current emission databases, underestimating the recent increase of CH4. Comparing the modified model output to the measurements, for which the annual global growth rates have been taken into account, the agreement over southern Africa and at some locations over Asia improves, but over northern Africa and on the flight route to northern South America, modeled mixing ratios become too high. This points to regionally different growth rates.
 Over tropical Asia, the amplitude of the seasonal cycle exceeds the interannual variations derived from global data sets from ground station networks. While on a global scale average interannual variations in CH4 mixing ratios are of the order of 5–10 ppb, CH4 varies by as much as 30–80 ppb within months over Asia. Over India, southeast China and the South China Sea maxima occur in August and September, when minimum value would expected due to high concentrations of OH radicals in summer which limit the lifetime of CH4. Focusing on four regions over tropical Asia, for which abundant data are available, the model is used to study the seasonality of CH4 in the upper troposphere with a tagged tracer simulation. This allows the analysis of the contributions from individual source regions to the observed seasonal cycles, being the result of the combined seasonality of emissions and transport processes. The simulation reproduces the general features of the seasonal cycle and, based on the regional tagging, disentangles the respective source region contributions. It is found that the emissions from China and South Asia have a dominant influence on the seasonal cycle over Asia. In addition, sources in Southeast Asia play an important role, especially over southeast China and the South China Sea. Over the Indian subcontinent even African emissions are important at times, though during the summer monsoon the foremost contribution is from the region itself, with a minor contribution from China.
 The upper tropospheric anticyclone, centered over northern India and Tibet, is the dominant meteorological system during the South Asian summer monsoon, and its development strongly influences the transport of emissions across tropical Asia, reaching out over the western Pacific. During the buildup of the anticyclone in spring and during its retreat in autumn, South Asian emissions are transported by westerly flows at its northern boundary, resulting in high mixing ratios over the western Pacific. When the anticyclone reaches its maximum strength in July and August, this outflow pattern is modified, and the transport toward the western Pacific and the South China Sea weakens. Emissions from biogenic sources, such as wetlands and rice paddies, are closely related to intense precipitation, while the reverse applies to those from open biomass burning. The Asian monsoon system can thus be considered the main driving force for the seasonality of upper tropospheric CH4 mixing ratios over Asia, because it influences the emission seasonality and dominates the air mass transport patterns.
 These findings confirm that aircraft data from cruise altitude contain a wealth of information about surface fluxes in regions where strong convection links the upper tropospheric mixing ratios to regional surface emissions. While some difficulties in the interpretation of these data arise from the fact that different flight routes were covered at different time periods, both data sets provide information for example to support satellite data validation and future modeling work such as inversion studies.
 We thank all CARIBIC partners as well as Lufthansa and Lufthansa Technik for their ongoing support. CARIBIC is also financially supported by Fraport AG. The CONTRAIL project is supported by Japan Airlines, the JAL foundation, and JAMCO. We especially acknowledge C. Koeppel, D. Scharffe, and S. Weber for operation of the CARIBIC container and K. Katsumata and H. Sandanbata for technical support of the CONTRAIL observations. T.J.S. was supported by JSPS/MEXT KAKENHI-A grant 22241008. We also would like to thank E. Dlugokencky for providing the most recent values of the global CH4 increase rate.