Aircraft observations of Asian outflow from the Transport and Chemical Evolution Over the Pacific (TRACE-P) aircraft mission over the NW Pacific (March and April 2001) show large CH4 enhancements relative to background, as well as strong CH4-C2H6-CO correlations that provide signatures of regional sources. We apply a global chemical transport model simulation of the CH4-C2H6-CO system for the TRACE-P period to interpret these observations in terms of CH4 sources and to explore in particular the unique constraints from the CH4-C2H6-CO correlations. We use as a priori a global CH4 source inventory constrained with National Oceanic and Atmospheric Administration (NOAA) Climate Monitoring and Diagnostics Laboratory (CMDL) surface observations [Wang et al., 2004]. We find that the observed CH4 concentration enhancements and CH4-C2H6-CO correlations in Asian outflow in TRACE-P are determined mainly by anthropogenic emissions from China and Eurasia (defined here as Europe and eastern Russia), with only little contribution from tropical sources (wetlands and biomass burning). The a priori inventory overestimates the observed CH4 enhancements and shows regionally variable biases for the CH4/C2H6 slope. The CH4/CO slopes are simulated without significant bias. Matching both the observed CH4 enhancements and the CH4-C2H6-CO slopes in Asian outflow requires increasing the east Asian anthropogenic source of CH4, and decreasing the Eurasian anthropogenic source, by at least 30% for both. The need to increase the east Asian source is driven by the underestimate of the CH4/C2H6 slope in boundary layer Chinese outflow. The Streets et al.  anthropogenic emission inventory for east Asia fits this constraint by increasing CH4 emissions from that region by 40% relative to the a priori, largely because of higher livestock and landfill source estimates. Eurasian sources (mostly European) then need to be reduced by 30–50% from the a priori value of 68 Tg yr−1. The decrease of European sources could result in part from recent mitigation of emissions from coal mining and landfills.
If you can't find a tool you're looking for, please click the link at the top of the page to go "Back to old version". We'll be adding more features regularly and your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Atmospheric methane (CH4) is an important greenhouse gas with an atmospheric lifetime of ∼10 years. The atmospheric abundance of CH4 has increased by a factor of 2.5 since preindustrial times [Etheridge et al., 1998]. The resulting radiative forcing is half that of CO2 [Hansen et al., 2000]. Methane also plays a central role in atmospheric chemistry. It is an important sink for tropospheric OH and hence affects the oxidizing power of the atmosphere. It is a major source of tropospheric ozone, with implications for both the oxidizing power of the atmosphere and air quality [Fiore et al., 2002]. Methane affects stratospheric ozone both by scavenging Cl radicals and by providing a source of water vapor. Better understanding of anthropogenic influences on CH4 is essential for addressing issues of climate change and global atmospheric chemistry.
 Considerable research has focused on quantifying sources of CH4 [Intergovernmental Panel on Climate Control (IPCC), 2001]. The major anthropogenic sources include rice cultivation, livestock, landfills, fossil fuel production and consumption (natural gas venting, leakage, and coal mining), and biomass burning. The major natural sources include wetlands and termites. The total emission from all sources is known to be in the range 460–600 Tg yr−1 because of independent constraints from halocarbon data on the global mean tropospheric OH concentration and hence on the CH4 sink [Prinn et al., 1995, 2001; Spivakovsky et al., 2000]. However, estimates for most of the individual source terms in the global CH4 budget have uncertainties of a factor of two or more according to IPCC . Better understanding of these individual source terms and their contributions to the long-term trend in CH4 is needed.
 The traditional approach for estimating CH4 sources has been by global extrapolation from limited flux measurements and socioeconomic data [e.g., Matthews and Fung, 1987; Matthews et al., 1991; Lerner et al., 1988; Cicerone and Oremland, 1988]. These bottom-up emission inventories suffer from substantial uncertainty due to undersampling of the sources and poor characterization of source variability. Global chemical transport model (CTM) simulations of atmospheric CH4 observations from the National Oceanic and Atmospheric Administration (NOAA) Climate Monitoring and Diagnostics Laboratory (CMDL) network of surface sites have been used to test the bottom-up emission inventories, either by a forward model with multiple scenarios [Fung et al., 1991] or by formal inverse modeling methods [Hein et al., 1997; Houweling et al., 1999; Wang et al., 2004; S. E. Mikaloff Fletcher et al., Constraining CH4 source estimates with atmospheric observations of CH4 and its 13C/12C isotopic ratios in CH4: 1. Inverse modeling of source processes, submitted to Global Biogeochemical Cycles, 2004a, hereinafter referred to as Mikaloff Fletcher et al., submitted manuscript, 2004a; S. E. Mikaloff Fletcher et al., Constraining CH4 source estimates with atmospheric observations of CH4 and its 13C/12C isotopic ratios in CH4: 2. Inverse modeling of fluxes from geographical regions, submitted to Global Biogeochemical Cycles, 2004b, hereinafter referred to as Mikaloff Fletcher et al., submitted manuscript, 2004b]. However, spatial overlap of the sources makes it difficult to derive a unique set of top-down constraints for CH4 sources from the limited surface air observations. Additional constraints on the CH4 budget and specific sources are provided by measurements of the carbon isotopes δ13C and 14CH4 [Lowe et al., 1991, 1994; Gupta et al., 1996; Quay et al., 1999; Miller et al., 2002; Mikaloff Fletcher et al., submitted manuscript, 2004a; Mikaloff Fletcher et al., submitted manuscript, 2004b], but these data are even more limited.
 We show here that improved top-down constraints on the sources of CH4 can be obtained from observations of CH4-C2H6-CO correlations interpreted with a global CTM. Ground-based and aircraft measurements show strong correlations of CH4 with C2H6 and CO [Harriss et al., 1994; Blake et al., 1996; Bartlett et al., 1996; Shipham et al., 1998; Matsueda and Inoue, 1999]. Ethane (C2H6) is a tracer of natural gas leakage, coal mining, and biomass burning [Rudolph, 1995]. CO is a general tracer of combustion. Both C2H6 and CO have atmospheric lifetimes of a few months, sufficiently long to serve as atmospheric tracers of their sources. One might therefore expect the biomass-burning source of CH4 to have a correlated CH4-C2H6-CO signature; the natural gas/oil/coal mining sources to have a correlated CH4-C2H6 signature; and the sources from livestock, rice cultivation, landfills, and wetlands to have no direct correlations with CO and C2H6. In practice, colocation of these different sources complicates the interpretation, and a CTM analysis is needed.
 We focus our analysis here on CH4-C2H6-CO correlations in Asian outflow, using observations from the NASA Transport and Chemical Evolution Over the Pacific (TRACE-P) aircraft mission over the western Pacific during March and April 2001 [Jacob et al., 2003]. Asia is of particular interest as a source of CH4 owing to its large population, rapid industrialization, and extensive rice cultivation. There is also considerable seasonal biomass burning in Southeast Asia in spring [Duncan et al., 2003]. The TRACE-P mission used two aircraft, based in Hong Kong and Japan, to characterize the chemical composition of Asian outflow. The major outflow pathways involved frontal lifting by warm conveyor belts (WCBs) ahead of southeastward moving cold fronts and boundary layer advection behind the cold fronts [Fuelberg et al., 2003; Liu et al., 2003]. Bartlett et al.  reported strong correlations of CH4 with C2H6, ethyne (C2H2), propane (C3H8), and carbon tetrachloride (C2Cl4) in the Asian outflow. They concluded that urban and combustion sources dominated the regional CH4 budget. However, a bottom-up inventory of Asian CH4 sources for 2000 generated by Streets et al.  in support of TRACE-P identified livestock, landfills, rice cultivation, coal mining, and biomass burning as the principal regional sources of CH4, in apparent contradiction with the analysis of Bartlett et al. . The Streets et al.  inventory also includes estimates of C2H6 and CO emissions, allowing a consistent interpretation of observed TRACE-P correlations using a CTM simulation.
 We use here the GEOS-CHEM CTM [Bey et al., 2001] in a consistent simulation of the CH4-C2H6-CO system for the TRACE-P period. This model has been used previously for global simulations of CH4 and CO aimed at understanding trends in the past decade [Wang et al., 2004; B. N. Duncan et al., Model study of the variability and trends of carbon monoxide (1988–1997): 1. Model formulation, evaluation, and sensitivity, submitted to Journal of Geophysical Research, 2004a; B. N. Duncan et al., Model study of the variability and trends of carbon monoxide (1988–1997): 2. Trends and regulating factors, submitted to Journal of Geophysical Research, 2004b]; Previous GEOS-CHEM applications to the analysis of TRACE-P data included investigation of Asian outflow pathways [Liu et al., 2003] and transpacific transport [Heald et al., 2003b; Jaeglé et al., 2003], as well as quantification of the Asian sources of CO [Palmer et al., 2003a; Heald et al., 2003a], CO2 (P. Suntharalingam et al., Improved quantification of Chinese carbon fluxes using CO2/CO correlations in Asian outflow, submitted to Journal of Geophysical Research, 2004, hereinafter referred to as Suntharalingam et al., submitted manuscript, 2004), nitriles [Li et al., 2003], halocarbons [Palmer et al., 2003b], and ozone [Liu et al., 2004]. Transport in the model is thus heavily diagnosed for the TRACE-P period, and there are no systematic biases [Kiley et al., 2003].
2. Model Description
2.1. General Description
 The GEOS-CHEM CTM (available at http://www.as.harvard.edu/chemistry/trop/geos/index.html, v4.26) is driven by assimilated meteorological fields from the Goddard Earth Observing System (GEOS) of the NASA Global Modeling and Assimilation Office (GMAO). We use GEOS-3 meteorological fields for the TRACE-P time period with a horizontal resolution of 1° × 1°, 48 vertical sigma levels up to 0.01hPa, and a temporal resolution of 6 hours (3 hours for surface variables and mixed layer depths). The horizontal resolution is degraded here to 2° latitude by 2.5° longitude for computational expediency.
 The main sinks of CH4, CO, and C2H6 are reactions with OH. The rate constants are from DeMore et al. . For CH4 we also include a minor sink from soil absorption (about 30 Tg yr−1) [Wang et al., 2004]. Global three-dimensional (3-D) monthly mean tropospheric OH concentrations from a full tropospheric oxidant simulation (GEOS-CHEM v4.33) are used [Fiore et al., 2003]. These OH concentrations yield an annual mean methyl chloroform (CH3CCl3) lifetime against loss by tropospheric OH of 6.3 years, which is within the range of 5.3–6.9 years estimated by Prinn et al.  from CH3CCl3 measurements. The same OH fields have been used in other GEOS-CHEM simulations of the TRACE-P data [Li et al., 2003; Heald et al., 2003a, 2003b; Jaeglé et al., 2003; Palmer et al., 2003a].
 Tagged tracer simulations for CH4, C2H6, and CO in GEOS-CHEM are conducted to quantify the contributions to atmospheric concentrations from different source regions and source types. These tagged tracer simulations all use the same OH concentrations as the standard simulation so that results are additive. Source regions include east Asia, Europe, eastern Russia, North America, and the rest of the world as defined in Figure 1. We will refer to “Eurasia” as the combination of Europe and eastern Russia. Source types for CH4 include livestock, rice cultivations, wetlands, biomass burning, biofuel, fossil fuel (gas/oil/coal mining), and landfills. The simulations of C2H6 and CO are initialized with a 2-year spin-up. The initialization of the CH4 simulation is described in section 2.3.
 We use the global 1998 inventory of Wang et al.  as our a priori estimate of CH4 emissions. The global CH4 growth rate was near zero between 1998 and 2001, presumably reflecting a leveling-off of emissions [Dlugokencky et al., 1998], although a positive trend of global OH concentrations could have contributed. Wang et al.  originally derived their inventory for 1994 by inverse analysis of the NOAA/CMDL surface observations, using GEOS-CHEM as the forward model, and then applied interannual variability to these emissions for 1988–1998 on the basis of socioeconomic and meteorological data. They found that they could capture in this manner much of the observed spatial and temporal variability of CH4 concentrations over the 1988–1997 decade.
Table 1 summarizes the global and east Asian CH4 sources from Wang et al.  and compares them with the global EDGAR3.2 emission inventory [Olivier and Berdowski, 2001] and to the east Asian emission inventory of Streets et al. . The latter two inventories include only anthropogenic sources and biomass burning. Anthropogenic sources here include rice cultivation, livestock, fossil fuel (gas/oil/coal mining), landfills, and biofuel. Classifying the biomass-burning source as natural or anthropogenic is somewhat subjective. For the purpose of this paper we will regard it as natural, largely out of convenience since its geographical distribution (mainly tropical) is distinct from that of the major anthropogenic sources. Annual east Asian emissions in the Wang et al.  inventory are dominated by wetlands, rice cultivation, livestock (including animal enteric fermentation and animal waste management), coal mining, and landfills. Natural gas and biofuel sources are relatively small. We refer to the sum of natural gas (production and transmission), oil (production, transmission, and handling), and coal mining as the fossil fuel source. The other fossil fuel sources of CH4 from industry, power generation, residential and commercial use, and transport are all minor [Olivier and Berdowski, 2001].
Table 1. East Asiana and Global CH4 Emissions (Tg yr−1)
These inventories include only anthropogenic sources and biomass burning.
Percentages in parentheses are the seasonal scaling factors for February to April 2001 (TRACE-P period) relative to the annual mean. For example, the value of 40% for rice cultivation means that the emissions in February to April are only 40% of the annual mean value.
 East Asian emissions from rice cultivation in the Streets et al.  inventory are 50% lower than in the work of Wang et al.  and the coal mining source is 40% lower, while livestock emissions are 30% higher and landfill emissions are a factor of two higher. The EDGAR3.2 inventory is better aligned with Streets et al. .
 Seasonal variation of CH4 emissions from rice cultivation, biomass burning, and wetlands is applied here following Wang et al. . Our focus is on the February to April (TRACE-P) period. East Asian emissions from rice cultivation peak in June to October (170% of annual mean flux) and are low during February to April (40% of annual mean flux). This is consistent with the Streets et al.  inventory, in which CH4 emissions from rice cultivation peak in fall and are low in spring. Emissions from natural wetlands in February to April are about 80% of the annual average. Biomass burning in Southeast Asia is by contrast at its seasonal high (240% of annual mean flux) during February to April.
Figure 2 shows the geographical distribution of CH4 emissions for February to April 2001 in the Wang et al.  inventory. There are important regional differences between sources. In China and India the major sources are livestock, coal mining, landfills, and rice cultivation. Landfills and gas/oil are the main sources in Japan. Wetlands, rice cultivation, biomass burning, and livestock are important sources in Southeast Asia. Fossil fuel (mostly from natural gas), landfills, and livestock are dominant in Eurasia.
 Major sources of C2H6 include natural gas leakage, natural gas venting, coal mining, biofuel use, and biomass burning [Rudolph, 1995]. We refer to the sum of the first three sources as the fossil fuel source. Anthropogenic emissions (fossil fuel and biofuel) for east Asia are taken from Streets et al. . Fossil fuel emissions for the rest of the world are scaled to the corresponding source of CH4, for which better a priori information is available. Biofuel emissions for the rest of the world are based on the Yevich and Logan  biofuel inventory with emission factors from Andreae and Merlet . Biomass burning emissions are scaled to the corresponding emission of CO [Duncan et al., 2003], with a C2H6/CO emission ratio dependent on fuel type [Staudt et al., 2003].
 Natural gas composition varies with geographical region [Rudolph, 1995]. In this work, CH4/C2H6 emission factors for the fossil fuel sources outside east Asia were derived by fitting the global C2H6 simulation to atmospheric concentration data including surface sites, ship cruises, aircraft missions, and total columns from ground-based remote sensing. We thus adopt molar CH4/C2H6 emission ratios from fossil fuel of 8 in eastern Russia, 24 in Europe, and 19 for the rest of the world except east Asia, where the use of Streets et al.  inventory for C2H6 yields a molar CH4/C2H6 emission ratio of 40. These values are highly variable but roughly in the range of U.S. energy industry data [Flores et al., 1999], which indicate emission ratios of 38 or higher (dry gas), 19–37 (gas condensate), and 11–18 (oil). Measurements in Chinese cities indicate CH4/C2H6 concentration ratios in the range of 5–35 (D. R. Blake, unpublished data, 2001). Our total global C2H6 source of 13.5 Tg yr−1 (Table 2) is consistent with previous literature estimates of 8–24 Tg yr−1 [Blake and Rowland, 1986; Kanakidou et al., 1991; Rudolph, 1995; Boissard et al., 1996; Gupta et al., 1998; Wang et al., 1998].
Table 2. CH4-C2H6-CO Emissions and Emission Ratios for February to April 2001a
Including Europe and eastern Russia (see Figure 1).
Including contributions from rice cultivation, livestock, fossil fuel, landfill, and biofuel emissions.
2.2.3. Carbon Monoxide
 Our CO simulation for the TRACE-P period is that of Palmer et al. [2003a], who conducted an inverse model analysis of Asian CO sources on the basis of the TRACE-P CO observations. They used a priori inventories from Streets et al.  for east Asian anthropogenic sources (fossil fuel and biofuel combustion), Heald et al. [2003a] for biomass burning in Southeast Asia, and Duncan et al.  for other sources. Transport was simulated with GEOS-CHEM, which served as the forward model for the inversion. Fitting the CO observations in TRACE-P required a 54% increase in anthropogenic emissions from China relative to the a priori, constrained by the observations in boundary layer outflow, and a 74% decrease in emissions from Southeast Asia (mostly from biomass burning), constrained by the observations in the free troposphere south of 30°N. Similar results for CO emissions were obtained by Heald et al. [2003b] from an analysis of Measurements of Pollution in the Troposphere (MOPITT) satellite observations of CO columns during the TRACE-P period. The optimized a priori sources of CO from Palmer et al. [2003a] are given in Table 2. The resulting simulation of CO concentrations, which we use here, is compared with TRACE-P observations in that paper. The mean bias is −4 ppb and the frequency distribution of differences is peaked around zero.
2.2.4. Emission Ratios
Table 2 gives the CH4-C2H6-CO emission ratios from selected sources, globally and for the two main source regions contributing to the structure of Asian outflow over the Pacific during TRACE-P: east Asia and Eurasia (Europe and eastern Russia). We exclude India from east Asia here, since India did not contribute significantly to the structure of Asian outflow during the TRACE-P period [Palmer et al., 2003a]. There is substantial spatial overlap between the different anthropogenic sources of CH4 (Figure 2); hence we also give for each region in Table 2 the lumped anthropogenic emission ratio.
 Within east Asia, the CH4/C2H6 emission ratio for fossil fuel (40 molar) is higher than that for biofuel (10) and close to that for biomass burning (35). The C2H6 emission factor for biofuel is much higher than for open biomass burning [Bertschi et al., 2003], apparently because of p ethane production during flaming combustion [Yokelson et al., 2003]. The CH4/CO emission ratio for fossil fuel (0.30 molar) is 2–3 times that for biofuel (0.10) and biomass burning (0.12). The C2H6/CO emission ratio from fossil fuel (7.6 × 10−3 molar) is lower than that for biofuel (10 × 10−3) but much higher than that for biomass burning (3.6 × 10−3). The CH4/CO and CH4/C2H6 emission ratios from all lumped anthropogenic sources are larger than those from the combustion sources owing to additional CH4 inputs from livestock, landfills, and rice cultivation.
 The CH4 and C2H6 anthropogenic emissions from Eurasia are larger than from east Asia while the CO source is much smaller, reflecting more vigorous pollution control of CO and greater natural gas production in Eurasia. There are thus large differences in the CH4/CO and C2H6/CO emission ratios for anthropogenic sources from east Asia versus Eurasia (0.57 versus 1.0 for CH4/CO, 8.8 × 10−3 versus 27 × 10−3 molar for C2H6/CO), and the CH4/C2H6 anthropogenic emission ratio is larger for east Asia (64 versus 37).
2.3. Initialization and Global Aspects of the CH4 Simulation
 Proper model initialization of the CH4 tagged tracers presents a difficulty because of the long lifetime of CH4. We wish to use these tracers to interpret the variability of CH4 in Asian outflow during March and April 2001. Because the timescale for mixing within the Northern Hemisphere is a few months, we expect the variability observed in TRACE-P to reflect fresh emissions from the past 2–3 months together with the latitudinal and vertical structure of the background. Accordingly, we define a background CH4 tracer in the simulation as the CH4 present in the atmosphere on 1 January 2001, including contributions from all sources. This background tracer has no emissions following 1 January and is allowed to decay by oxidation by OH. The other tagged tracers, representing different CH4 source regions and source types, are initialized with zero concentration on 1 January. They build up over the course of the simulation as a result of emissions and are removed by oxidation by OH. The same archived OH concentration fields are used to calculate the losses of all tracers; hence the sum of the individual tracers adds up to the total CH4 concentration.
 We need to define a global field for the background tracer on 1 January 2001. We start from NOAA/CMDL global surface observations for December 2000 (available at ftp://ftp.cmdl.noaa.gov) [Dlugokencky et al., 1994, 2001] to define the latitudinal gradient of CH4 concentrations at the surface. From this initial state with no vertical gradients in the troposphere, we conduct a 2-year spin-up simulation with 2000 meteorological fields and the Wang et al.  emissions to construct the vertical and longitudinal gradients of CH4. The results from this 2-year spin-up are then scaled globally with a uniform factor to optimize agreement with the NOAA/CMDL surface data for December 2000.
 Evaluation of the model background for the TRACE-P period is shown in Figure 3 by comparison of model results for March 2001 with observations from NOAA/CMDL sites. The model reproduces the latitudinal gradient in the background without bias over the range of interest (10–50°N), except at a few continental sites around 45°N (LEF, HUN) where observed concentrations are high due to local source influence. Simulation of vertical gradients is discussed below using the TRACE-P observations.
3. Model Simulation of TRACE-P Observations
 We first examine the ability of the model driven by a priori sources from the Wang et al.  inventory (Table 1) to reproduce the general features of the CH4 distributions and CH4-C2H6-CO correlations observed in TRACE-P. The TRACE-P CH4 measurements rely on the same calibration standards as NOAA/CMDL so that CH4 concentrations should be consistent across the two data sets. We will show also results from a simulation where the Streets et al.  anthropogenic emissions for CH4 in east Asia have been superimposed on the Wang et al.  inventory. The ability of the model to simulate C2H6 observations using the sources in Table 2 will also be examined. A similar evaluation for CO, also using the sources of Table 2, was presented by Palmer et al. [2003a] and is not reported here. Throughout this paper, the model values are sampled along the flight tracks and the observations are averaged over model grid boxes. Stratospheric data (O3 > 100 ppb with CO < 100 ppb) are excluded. The observations are averaged over model grid boxes, corresponding to an averaging time span of 10–30 min for typically aircraft speeds. The model results are output every 3 hours. Considering that we are looking at aged air masses, at least 1 day downwind of the sources, the time discrepancy between the model and observation should be negligible. Linear regressions use the RMA (reduced major axis) method [Hirsch and Gilroy, 1984], accounting for errors in both variables.
Figure 4a (top) compares the simulated and observed CH4 concentrations for the ensemble of TRACE-P observations over the NW Pacific (west of 160°E). The simulation shows strong correlation with observations (r2 = 0.71). The 6% positive bias (slope of the RMA regression line) is driven by boundary layer Asian outflow rather than by the background. This structure in the bias is more apparent in Figure 4b, which shows the simulated versus observed mean vertical gradients of CH4 concentrations in TRACE-P west of 160°E. Model results using the Streets et al.  east Asian inventory are also shown. We distinguish between data north of 30°N, where anthropogenic outflow was strongest, and south of 30°N, where most of the Southeast Asian biomass burning influence was observed [Palmer et al., 2003a]. The model results with a priori sources show positive bias in the simulation of the Asian outflow enhancement at 0–6 km altitude, particularly north of 30°N. The bias is larger when the Streets et al.  inventory is used owing to higher livestock and landfill emissions in that inventory. Rice emissions are lower in the Streets et al.  inventory than in the work of Wang et al. , resulting in total annual east Asian CH4 emissions that are similar between the two inventories (Table 1). However, the rice emissions are at low latitudes (Figure 2) and are also at their seasonal minimum in February to April, thus making little contribution to the Asian outflow observed in TRACE-P. Although it would seem from Figure 4b that the Streets et al.  inventory degrades the simulation, we will show in section 4 that using that inventory in combination with a decrease in Eurasian emissions provides in fact a good fit to CH4 concentrations for the TRACE-P period.
Figure 4c (left) compares simulated and observed CH4 concentrations west of 160°E as a function of latitude and at different altitudes. Also shown (dashed lines) is the contribution of the CH4 background (emitted before 1 January 2001) to the model fields. The model reproduces the general increasing trend of CH4 with latitude, reflecting both the background and the Asian outflow enhancements. The model underestimates observations south of 16°N, but these are mainly from a single flight over the South China Sea (Figure 1), which makes interpretation difficult. CH4 enhancements above the background in the model fields (difference between solid and dashed lines) reflect emissions over the January to April 2001 period. As shown in Figure 4c (right), we find that east Asian sources contribute most of this enhancement south of 35°N, while Eurasian sources (Europe and eastern Russia; the former is about 80%) dominate north of 35°N. The principal east Asian sources contributing to the enhancements below 4 km are livestock, landfills, and coal mining. The principal Eurasian sources contributing to the Asian outflow are natural gas, landfills, and livestock. Biomass burning in Southeast Asia is a relatively large source south of 30°N, mostly above 8 km (convective outflow), and at 2–4 km (WCB frontal outflow), consistent with TRACE-P observations for the HCN and CH3CN biomass burning tracers [Li et al., 2003; Singh et al., 2003]. The wetlands source of CH4 does not contribute significantly to the Asian outflow sampled in TRACE-P because it is located too far south (see Figure 2).
 Comparisons of simulated and observed C2H6 concentrations are shown in Figure 5a (scatterplot), Figure 5b (vertical profiles), and Figure 5c (latitudinal gradients). There is strong correlation in the ensemble of data (r2 = 0.73). The regression slope of 1.04 in the scatterplot (Figure 5a) is mainly driven by the underestimate of the background. The magnitude of the enhancement in Asian outflow is well described, as shown in Figure 5b. The latitudinal gradients are similar to those of CH4, and there is no distinct bias in the model for any of the major features. The Eurasian fossil fuel source makes the largest contribution to the model fields at all altitudes north of 30°N. East Asian sources make only a minor contribution north of 30°N but are relatively more important to the south, where the Eurasian influence is less.
3.3. CH4-C2H6-CO Correlations
 We saw in section 3.1 that the model with a priori sources reproduces the general latitudinal and vertical distributions of CH4 in TRACE-P but is too high in the lower tropospheric Asian outflow. We will now examine what additional insights can be obtained from the observed CH4-C2H6-CO correlations. Observed pollution plumes with CO concentrations greater than 325 ppb on a 5-min average basis (4% of the data in all) are removed from the correlation analysis because they would otherwise overly influence the least squares statistics. We employ the bootstrap method [Venables and Ripley, 1999] to calculate the standard deviations of the slopes of the regression lines.
Figure 6 compares the simulated and observed CH4-C2H6-CO correlations for regional subsets of the TRACE-P data. The slopes of the regression lines and associated standard deviations are given in Table 3. We will use the CH4-C2H6 and CH4-CO correlations as constraints on CH4 sources, and the C2H6-CO correlation to test the relative influence of east Asian and Eurasian sources. The region north of 30°N and below 2 km (“Chinese outflow”) sampled anthropogenic outflow but was largely devoid of biomass burning influence [Liu et al., 2003; Palmer et al., 2003a; Russo et al., 2003; Singh et al., 2003]. The region south of 30°N and below 2 km (“tropical Asian outflow”) has more influence of sources from the tropical Asian region. In order to obtain independent information on Japanese/Korean sources, we use data from DC-8 flight 17 (31 March), which sampled a mix of Japanese and Korean influences but with relatively little Chinese influence [Blake et al., 2003; Suntharalingam et al., submitted manuscript, 2004]. Finally, we also examine the correlations in background air (east of 160°E and 2–8 km altitude).
Table 3. Slopes of CH4-C2H6-CO Relationships in Asian Outflow Over the NW Pacific (March to April 2001)a
Chinese Outflow (>30°N, 0–2 km)
Tropical Asian Outflow (20–30°N, 0–2 km)
Background (2–8 km, East of 160°E)
Slopes of the reduced-major-axis (RMA) regressions for the regional subsets of data shown in Figure 6, with standard deviations computed by the bootstrap method (see text). Units are mol/mol for CH4/C2H6 and CH4/CO, 10−3 mol/mol for C2H6/CO.
Results shown for simulations with different CH4 sources. For all scenarios the CO simulation is from Palmer et al. [2003a], in which Asian CO sources have been optimized through inversion analysis of the TRACE-P data, and C2H6 sources are as described in section 2.2.2.
 For Chinese outflow (Figure 6a and Table 3), the simulated CH4/CO and C2H6/CO slopes are consistent with observations, while the CH4/C2H6 slope is significantly too low (39 ± 1.1 versus 43 ± 1.2 × 10−3 molar). The CH4 and C2H6 concentrations in Chinese outflow reflect contributions from anthropogenic sources in both east Asia and Eurasia (section 3.1 and 3.2). Correction of the CH4/C2H6 bias will be discussed in section 4.
 For tropical Asian outflow (Figure 6b) the observed CH4/CO and C2H6/CO slopes are higher than for Chinese outflow (0.46 versus 0.38 for CH4/CO, 11 versus 8.5 × 10−3 molar for C2H6/CO) while the CH4/C2H6 slopes are similar (42 versus 43). As shown by the tagged CH4 simulation, tropical CH4 sources from wetlands and biomass burning are of little importance and anthropogenic sources of CH4 from east Asia and Eurasia remain dominant in this region. The model captures the regional differences in the CH4/CO and C2H6/CO slopes. However, it overestimates the CH4/C2H6 and CH4/CO slopes significantly.
 Compared with the Chinese and tropical Asian outflow, the Japanese/Korean plumes (Figure 6c) feature significantly higher CH4/C2H6 slopes (45 versus 42–43), CH4/CO (0.65 versus 0.38–0.46), and C2H6/CO (14 versus 9–11 × 10−3 molar). These regional differences are due to CO emission controls and to low biofuel emissions (characterized by a low CH4/C2H6 emission ratio). Again, the model captures this regional variation. It overestimates the CH4/C2H6 slope and underestimates the C2H6/CO slope, with no significant bias in the CH4/CO slope (Table 3).
Figure 6d shows the CH4-C2H6-CO correlations in background air masses east of 160°E. The observed slopes are much higher (58 for CH4/C2H6, 0.75 for CH4/CO, and 13 × 10−3 molar for C2H6/CO) than in Asian outflow. These correlations reflect the common large-scale latitudinal gradients of the three gases. The model reproduces all these correlations without significant bias.
4. Optimization of CH4 Sources
 We have shown that both east Asian and Eurasian anthropogenic sources were major contributors to the CH4 enhancement in Asian outflow observed in TRACE-P and that the correlations of CH4 with C2H6 and CO have distinct characteristics related to both regional sources and background. We seek now to use the constraints from the observed CH4 enhancements and CH4-C2H6-CO correlations to optimize the representation of CH4 sources in the model in a forward fashion. Considering that the C2H6 and CO simulations match the observations closely (see section 3.2 and Palmer et al. [2003a]), we focus here on adjusting CH4 emissions only. Since the east Asian and Eurasian anthropogenic sources are two dominant components affecting the structure of CH4 Asian outflow during TRACE-P, we choose to fit the model to the observations by adjusting these two CH4 emission variables. We do not try to resolve the individual anthropogenic source components of CH4 within each region because of strong spatial source overlap (Figure 2), which precludes detection of individual signatures in the outflow.
 We focus on the data for Chinese outflow (Figure 6a), which provide the strongest regional constraints on sources and for which model transport errors are smallest because no vertical motions are involved [Kiley et al., 2003]. The model with a priori sources overestimates the CH4 enhancement in this data set by 7 ppb (Figure 4b). The observed CH4/C2H6 and CH4/CO slopes are 43 ± 1.2 mol mol−1 and 0.38 ± 0.02 mol mol−1, respectively, while the corresponding model slopes are 39 ± 1.1 mol mol−1 and 0.38 ± 0.02 mol mol−1 (Table 3). We seek to satisfy the observed CH4 enhancements and the observed CH4/C2H6 and CH4/CO correlations by adjusting CH4 anthropogenic sources from east Asia and Eurasia in the model. Figure 7 shows the simulated CH4/C2H6 and CH4/CO slopes and the difference between simulated and observed mean CH4 concentration enhancements in the Chinese boundary layer outflow as a function of percentage changes in these two emission variables relative to the a priori. We aim to fit the observed mean CH4 enhancements to within 5 ppb and the slopes to within the 95% confidence interval from the bootstrap method (±1.96σ), i.e., 40–46 for CH4/C2H6 and 0.34–0.42 for CH4/CO. These ranges are shown as green and black lines in Figure 7. The domain of model space satisfying the three constraints is shaded in Figure 7. It defines our range of optimized east Asian and Eurasian anthropogenic emissions.
 The model simulation with a priori sources is represented by point A in Figure 7. The positive model bias in simulating the CH4 concentration enhancement in Asian outflow could be caused by an overestimate of either east Asian or Eurasian sources. We see from Figure 7 that simply decreasing either of these (or both) would cause the CH4/C2H6 slope to be too low. Fitting the CH4/C2H6 constraint requires an increase in east Asian emissions, and thus our solution involves increasing the east Asian and decreasing the Eurasian anthropogenic emissions both by at least 30% (point C).
 A 40% increase in springtime east Asian anthropogenic emissions relative to the a priori Wang et al.  inventory can be achieved simply by using instead the emission inventory of Streets et al.  (Table 1). Although that inventory is comparable to Wang et al.  on an annual basis, it is 40% higher in March and April when the rice paddy source is at its seasonal minimum, as discussed in section 2. Such a 40% increase of east Asian sources requires from Figure 7 a corresponding decrease in Eurasian sources of 30–50%.
Table 3 gives the simulated CH4/C2H6 and CH4/CO slopes in the different TRACE-P regional data subsets for (1) the a priori simulation, (2) the a priori but with Streets et al.  east Asian anthropogenic sources, (3) the a priori but with Eurasian anthropogenic sources decreased by 30–50%, and (4) the combination of cases 2 and 3. Using the Streets et al.  emission inventory increases the CH4/C2H6 and CH4/CO slopes in Chinese outflow, while decreasing the Eurasian sources decreases the slopes. Results with our optimized case 4 emissions indicate a general improvement in the simulation of the observed correlations for the tropical Asian as well as the Chinese boundary layer outflow, while there is little effect on the Japanese/Korean and background data subsets. The effects on the simulations of the CH4 vertical profiles and the CH4-C2H6-CO correlations are shown in Figures 4a–4c and 6. The ability of the model to fit the observations is clearly improved; only the CH4/CO slopes in the Chinese outflow and Japanese/Korean outflow are slightly worse. With the optimized CH4 sources, the simulated CH4 concentrations in March 2001 at CMDL surface sites north of 30°N are 7 ppb lower on average than with the a priori simulation (Figure 3), while there is little difference south of 30°N.
 Our a priori total of 68 Tg yr−1 for the Eurasian anthropogenic source from Wang et al.  is comparable to that in the EDGAR3.2 [Olivier and Berdowski, 2001] bottom-up inventory (64 Tg yr−1). These inventories were constructed for the years 1998 and 1995, respectively. Aside from the intrinsic uncertainties in emission estimates, it is also likely that Eurasian emissions have decreased in recent years owing to increased mitigation efforts. Such mitigation efforts involve, for example, capture and utilization of CH4 from landfills and coal mining and a shift from underground to surface mining of coal [Olivier, 2002a, 2002b; Environmental Protection Agency (EPA), 2001]. Our results suggest that large decreases in Eurasian emissions of CH4 could have taken place over the past decade.
 Large CH4 enhancements above background were observed in Asian outflow over the NW Pacific during the TRACE-P aircraft mission (March and April, 2001). The CH4 enhancements showed strong correlations with C2H6 and CO that varied with the outflow region. We used a global chemical transport model (GEOS-CHEM) simulation of the CH4-C2H6-CO system for the TRACE-P period to interpret these observations with the goal of better quantifying CH4 regional sources.
 Our a priori model simulation uses the Wang et al.  global CH4 emission inventory previously derived from inverse modeling of NOAA/CMDL observations. It reproduces the general latitudinal and vertical distributions of CH4 observed during TRACE-P but overestimates by about 7 ppb the CH4 enhancement in Asian outflow. This enhancement is driven by anthropogenic sources in both east Asia and Eurasia (defined here as Europe and eastern Russia). The a priori model also underestimates the CH4/C2H6 slopes in Chinese outflow and overestimates them in tropical Asian outflow and in Japanese/Korean plumes. It reproduces the observed CH4/CO slopes without significant bias except for an overestimate in tropical Asian outflow. We find that the CH4-C2H6-CO correlations in the Asian outflow observed in TRACE-P are defined mainly by anthropogenic sources of CH4 (landfills, livestock, and fossil fuel) from east Asia and Eurasia, rather than by biomass burning or wetland sources from tropical Southeast Asia. Emissions from rice cultivation did not make a significant contribution to the CH4 outflow in TRACE-P, in part because they are at low latitudes and in part because they were at their seasonal minimum during the TRACE-P period. Observations in late summer or early fall would be expected to see a stronger influence from rice cultivation CH4 emissions.
 We employ the constraints offered by the observed CH4 enhancements and the CH4-C2H6-CO correlations to optimize our estimates of the CH4 anthropogenic emissions in east Asia and Eurasia. To correct the positive model bias of CH4 concentrations in the boundary layer Chinese outflow while increasing the CH4/C2H6 slope and maintaining the CH4/CO slope, we find that we need to increase the east Asian and decrease the Eurasian anthropogenic sources of CH4 by at least 30% for both. The increase in east Asian emissions is imposed by the constraint on the CH4/C2H6 slope. Without this increase, the model slope would be too low.
 A 40% increase in east Asian anthropogenic emissions for the TRACE-P period relative to the Wang et al.  inventory can be implemented simply by substituting the Streets et al.  regional inventory for east Asia prepared in support of the TRACE-P mission. The Streets et al.  inventory has higher livestock and landfill emissions. The Eurasian anthropogenic source (80% European) then needs to be reduced by 30–50% from the Wang et al.  inventory. Decrease of European CH4 emissions could have taken place over the past decade owing to mitigation efforts. Intrinsic uncertainties in the a priori emission inventory must also be considered.
 We have shown in this work that observed CH4-C2H6-CO correlations in continental outflow place valuable constraints on the CH4 budget. In our application these constraints allowed us to distinguish between east Asian and Eurasian source signatures in Asian outflow over the Pacific and to conclude that the overestimate of CH4 in our simulation with a priori sources had to be corrected by both decreasing the Eurasian source and (remarkably) increasing the east Asian source. A more formal inverse modeling approach for the CH4-C2H6-CO system should be applied in a next stage to better exploit the constraints from the correlations. Improved understanding of C2H6 sources (adjusted here in a top-down fashion to match observations) will be needed in order to better relate CH4-C2H6 correlations to the fossil fuel source of CH4.
 We thank Qinbin Li, Hongyu Liu, Mathew Evans, and John Lin for helpful discussions. This work was funded by the NASA Carbon Cycle Program and the NSF Atmospheric Chemistry Program.