transcom model simulations of methane: comparison of

1

TransCom model simulations of methane: comparison of vertical 1

profiles with in situ aircraft measurements 2

3 Ryu Saito1, Prabir K. Patra1,2, Colm Sweeney3, Toshinobu Machida4, Maarten Krol5,6, Sander 4 Houweling6, Philippe Bousquet7, Anna Agusti-Panareda8, Dmitry Belikov4, Dan Bergmann9, 5 Huisheng Bian10, Philip Cameron-Smith9, Martyn P. Chipperfield11, Audrey Fortems-Cheiney7, 6 Annemarie Fraser12, Luciana V. Gatti13, Emanuel Gloor11, Peter Hess14, Stephan R. Kawa10, 7 Rachel M. Law15, Robin Locatelli7, Zoe Loh15, Shamil Maksyutov4, Lei Meng16, John B. 8 Miller3, Paul I. Palmer12, Ronald G. Prinn17, Matthew Rigby17,18, Christopher Wilson11 9 10 1. Research Institute for Global Change/JAMSTEC, 3173-25 Show-machi, Yokohama, 236-0001, Japan 11 2. Center for Atmospheric and Oceanic Studies, Tohoku University, Sendai, Japan 12 3. NOAA Earth Systems Research Laboratory, Boulder, CO, USA. 13 4. CGER, National Institute for Environmental Studies, 16-2 Onogawa, Tsukuba, 305-8506, Japan 14 5. Wageningen University and Research Centre, Droevendaalsesteeg 4, 6708 PB Wageningen, The 15

Netherlands 16 6. SRON Netherlands Institute for Space Research, Sorbonnelaan 2, 3584 CA Utrecht, The Netherlands 17 7. Universite de Versailles Saint Quentin en Yvelines (UVSQ), GIF sur Yvette, France 18 8. European Centre for Medium-Range Weather Forecasts (ECMWF), Reading, UK 19 9. Atmospheric, Earth, and Energy Division, Lawrence Livermore National Laboratory, 7000 East 20

Avenue, Livermore, CA94550, USA 21 10. Goddard Earth Sciences and Technology Center, NASA Goddard Space Flight Center, Code 613.3, 22

Greenbelt, MD 20771, USA 23 11. Institute for Climate and Atmospheric Science, School of Earth and Environment, University of 24

Leeds, Leeds, LS2 9JT, UK 25 12. School of GeoSciences, University of Edinburgh, King's Buildings, West Mains Road, Edinburgh, 26

EH9 3JN, United Kingdom 27 13. Divisao de Quimica Ambiental, Insituto de Pesquisas Energeticas e Nucleares, Sao Paulo, Brasil 28 14. Cornell University, 2140 Snee Hall, Ithaca, NY 14850, USA 29 15. Centre for Australian Weather and Climate Research, CSIRO Marine and Atmospheric Research, 30

107-121 Station St., Aspendale, VIC 3195, Australia 31 16. Dept. Geography and Environ. Studies Prog., Western Michigan Univ., Kalamazoo, MI 49008, USA 32 17. Center for Global Change Science, Building 54, Massachusetts Institute of Technology, Cambridge, 33

MA, 02139, USA 34 18. School of Chemistry, University of Bristol, Bristol, UK 35

SAITO ET AL.: MODEL-‐OBSERVATION COMPARISON OF METHANE VERTICAL PROFILES

For submission to J. Geophys. Res. – Atmos. D R A F T 9/28/12 11:18 PM

2

36

Abstract. 37

To assess horizontal and vertical transports of methane (CH4) concentrations at different 38

heights within the troposphere, we analyzed a subset of 12 chemistry-transport model 39

(CTM) simulations from the TransCom-CH4 intercomparison experiment. Model results 40

are compared with in situ aircraft measurements at 13 locations in Amazon/Brazil, 41

Mongolia, Pacific Ocean, Siberia/Russia and United States during the period of 2001 to 42

2007. The simulations generally show good agreement with observations for seasonal 43

cycles and vertical gradients. The correlation coefficients of the daily averaged model 44

and observed CH4 time series for the analyzed years are generally larger than 0.5, and 45

the observed seasonal cycle amplitudes are simulated well at most sites, considering the 46

between-model variances. However, larger deviations show up below 2 km for the 47

model-observation differences in vertical profiles at some locations, e.g., at Santarem, 48

Brazil, and in the upper troposphere and lower stratosphere region, e.g., at Surgut. 49

Vertical gradients are underestimated at Southern Great Planes, United States and 50

overestimated at Surgut, Russia. Systematic over- and under-estimation of vertical 51

gradient can be explained by inaccurate emission, whereas the varying results found for 52

the Brazilian Amazon are likely caused by problems in simulating vertical transport. 53

Overall, the zonal and latitudinal variations in CH4 are controlled by surface emissions 54

below 2.5 km, and transport patterns in the middle and upper troposphere. We show that 55

the models with larger vertical gradients, coupled with slower horizontal transport, 56

exhibit greater CH4 interhemispheric gradients in the lower troposphere. These findings 57

have significant implications for the future development of more accurate CTMs with 58

the possibility of reducing biases in estimated surface fluxes by inverse modeling. 59

60



3

1. Introduction 61

Methane (CH4) is an important greenhouse gas, contributes to photochemical smog 62

formation under polluted conditions by reaction with hydroxyl radical (OH), and in the 63

stratosphere CH4 is the main source of water vapour [Crutzen and Graedel, 1993]. Over 64

the last two decades, CH4 simulations have been conducted using 3-dimensional global 65

chemistry transport models (CTMs). Recently, a model inter-comparison experiment 66

(TransCom-CH4) was organized with the aim to assess the role of transport model 67

uncertainties in simulating global CH4. The first analysis [Patra et al., 2011] 68

concentrated on background concentrations at the surface, and showed overall 69

consistent results across the participating models. However, that analysis was limited to 70

8 surface sites, where simultaneous measurements of CH4, sulphur hexafluoride (SF6) 71

and methyl chloroform (CH3CCl3) were available for more than 5 years in the period 72

1990-2007. Patra et al. [2011] highlighted the importance of inter-hemispheric (IH) 73

transport, and demonstrated that the models consistently under or overestimate the 74

latitudinal gradients of all the three species depending on the simulated rate of 75

inter-hemispheric mixing. How the simulated gradient varies with altitude, however, has 76

not been studied in detail yet [e.g., Nakazawa et al., 1991]. 77

78

The TransCom-CH4 database includes simulated vertical profiles, which can be 79

compared to data from several aircraft measurement campaigns that have been carried 80

out during the last decade [Tans et al., 1996; Machida et al., 2001; Gatti et al., 2010]. 81

The zonal mean simulation results show that CH4 concentrations, within the 82

troposphere, decrease with altitude in the northern hemisphere (NH) and increase with 83

altitude in the southern hemisphere (SH) [Patra et al., 2011]. Both the vertical and 84

meridional gradients exhibit seasonal dependency in relation to seasonality in transport, 85

surface emissions or chemical loss. However this conjecture is based only on the model 86

intercomparison, without comparison with observations. Thus, an evaluation of the 87



4

modeling results with in situ observations will likely improve our understanding of CH4 88

transport mechanisms of vertical and interhemispheric exchange in the troposphere. 89

90

The purpose of this study is to evaluate the simulated CH4 concentrations using airborne 91

measurements. The model-observation difference depends on uncertainties in surface 92

fluxes, model transport and chemistry. We assess the flux and transport dependencies of 93

the model-observation differences using vertical profiles over a variety of surface 94

conditions and transport regimes, e.g., wetlands in Siberia, tropical forests in Amazon, 95

boreal forests of North America and the marine environment. In the next sections, we 96

briefly describe the model simulation scenarios and the aircraft measurement programs, 97

followed by results and discussions. Some conclusions are given in Section 4. 98

99

100

2. Models and Measurements 101

2.1. Simulation concentrations 102

The institutes participating in the TransCom-CH4 experiment produced global CH4 103

distribution using 13 models and 4 variants with different spatial resolutions and 104

hydroxyl radical (OH) distributions. The CTMs simulation outputs include CH4 vertical 105

profiles over various sites for a decadal period (1992−2007) with a 1 or 3 hr time 106

interval, for six CH4 tracers that used different surface emission scenarios. A detailed 107

description of the TransCom-CH4 experiment is documented in Patra et al. (2011). In 108

this study, we use CH4 profiles simulated by 12 CTMs (Table 1), namely the ACTM 109

[atmospheric general circulation model (AGCM)-based CTM; Patra et al., 2009], CAM 110

[Community Atmosphere Model; Gent et al., 2009], CCAM [Conformal-Cubic 111

Atmospheric Model; Law et al., 2006], GEOS-Chem [Goddard Earth Observing System 112

(GEOS) meteorology driven CTM; Fraser et al., 2011], IFS [Integrated Forecasting 113

System; www.ecmwf.int/research/ifsdocs/CY36r1/index.html], IMPACT_1x1.25 114



5

[Integrated Massively Parallel Atmospheric Chemical Transport; Rotman et al., 2004], 115

LMDZ [Laboratoire de Météorologie Dynamique – Zoom; version 4; Hourdin et al., 116

2006], MOZART [Model for Ozone and Related Chemical Tracers; version 4; Emmons 117

et al., 2010], NIES08i [NIES CTM; version 2008 isentropic; Belikov et al., 2011], 118

PCTM [Parameterized CTM; Kawa et al., 2004], TM5_1x1 [Transport Model; version 119

5; Krol et al., 2005] and TOMCAT CTM [Chipperfield, 2006]. Five models, ACCESS 120

[The Australian Community Climate and Earth-System Simulator; Corbin and Law, 121

2011] driven by climatological meteorology, ACTM_OH (OH sensitivity run of 122

ACTM), GEOS-Chem_DOH (OH sensitivity run), IMPACT (higher resolution version 123

used), and TM5 (higher resolution version used), are excluded from main discussions of 124

this analysis. Two of the models (ACTM and GEOS-Chem) repeated their TransCom 125

runs with the same resolution and meteorology but with different OH fields. For both 126

models, the vertical profile gradient was similar in the two different runs. The IMPACT 127

and TM5 models repeated their TransCom runs using different horizontal resolution 128

(implicitly different meteorology). For both models we find relatively large differences, 129

greater than the differences between different models, in some seasons. No clear 130

distinction can be made whether the higher or lower resolution version of the models is 131

better for these two models (Fig. S1). However, it has been shown earlier that the higher 132

resolution version of TM5 performs better than its coarser version for the 133

interhemispheric gradients of SF6, CH4 and CH3CCl3 near the Earth’s surface [Patra et 134

al., 2011]. 135

136

We used CH4 tracers simulated using (1) the control emission scenario (CTL) (Figure 1) 137

that combined seasonally varying natural sources [CYC NAT; major components being 138

the emissions from biomass burning and wetlands; see Fung et al., 1992], interannually 139

varying anthropogenic emissions [ANT_IAV; based on EDGAR3.2; see Olivier and 140

Berdowski, 2002 and Patra et al., 2009 for details], and (2) the EXTRA flux, in which 141



6

interannually varying biomass burning emissions are based on GFEDv3 [van der Werf 142

et al., 2006] and wetland emissions are from the VISIT terrestrial ecosystem model [Ito 143

and Inatomi, 2012]. Based on our experiences in Patra et al. [2011], these two tracers 144

are sufficient to evaluate the differences in the simulations caused by the emission 145

distribution and strength. To remove systematic offsets between the CTM simulations 146

the decadal mean concentrations in the altitude range of 2−6 km over Hawaii (HAA) is 147

subtracted from the results. For all statistical calculations, the model values are sampled 148

at the time and location of the observations. The simulations of atmospheric 222Radon 149

and SF6 concentration are used for checking some of the model transport properties. 150

Detailed analysis of 222Radon simulations has been done for vertical profiles in relation 151

to deep cumulus convection and surface sites [Belikov et al., 2012]. 152

153

2.2. Airborne in situ measurements 154

Vertical CH4 profiles have been measured in small aircraft by the NOAA ESRL at sites 155

in North America and the Pacific 156

(http://www.esrl.noaa.gov/gmd/ccgg/aircraft/index.html), by NIES at sites in Siberia 157

[Machida et al., 2001] and by IPEN at sites in Amazon [Gatti et al., 2010]. The site 158

names are listed in Table 2 and their locations are plotted with blue symbols on a global 159

CH4 emission map in Figure 1. The aircrafts measured at altitudes up to 5 to 10 km 160

dependent on the site (Table 2). The airborne measurements are sampled a few times a 161

month with measurements continuing to the present for all sites except for Amazon 162

(SAN) where they terminated in 2005. The observation period for each site overlapping 163

with our analysis period is shown in Table 2. 164

165

The samples collected over Surgut, Russia are analyzed using gas chromatography 166

(Agilent 5890, Agilent Technologies Inc.) equipped with a flame ionization detector 167

(GC-FID) calibrated against the NIES-94 CH4 scale. The analytical precision of the 168



7

GC-FID for the CH4 concentration was estimated to be less than 2.0 ppb. The NIES-94 169

CH4 scale is higher than the WMO (NOAA) scale by 3.5–4.6 ppb in the concentration 170

range of 1750–1840 ppb [Zhou et al., 2009]. 171

172

The NOAA ESRL aircraft measurements are made by collecting air samples in 0.7 L 173

silicate glass flasks at biweekly to monthly intervals at about 16 sites on regular basis 174

and on campaign mode over Santarem, Brazil (Table 2). Air samples are collected using 175

a turboprop aircraft in the altitude range of about 500 m to 8 km. CH4 is measured by 176

gas chromatography (GC) equipped with flame ionization detection at accuracy of ±1.2 177

ppb. The concentrations are reported at the absolute calibration scale of NOAA that is 178

maintained as the WMO standard [Dlugokencky et al., 2005]. 179

180

181

3. Results and discussions 182

We compare the CTM simulation results with the airborne in situ measurements to 183

evaluate the differences between the models and models with measurements during the 184

analysis period of 2001 to 2007. 185

186

3.1. Seasonal and daily variations at different heights 187

Figure 2 shows the observed and simulated time series of CH4 concentration for 2007 188

for 3 different height ranges (0−2 km, 3−5 km and 6−8 km) and six sites. Information 189

on seasonal cycles is sometimes difficult to extract from the measurements given the 190

irregularity in data coverage and gaps in the time series (sky-blue dots). For better 191

clarity, multi-model mean seasonal cycles (black line) are also shown, which are created 192

by fitting a curve (3 harmonics) to the afternoon (1200-1600 local time) mean time 193

series of all the modeling results. This method ensures unambiguous identification of 194

the peaks and troughs in the simulated seasonal cycles of CH4, which is otherwise the 195



8

case for large data gaps at some of the observation sites. The amplitude and phase of the 196

simulated seasonal cycles is similar for all models. Although, the synoptic and daily 197

scale variations show differences between the modeling results, we find the observed 198

seasonal cycle is matches models to better than the 1-σ standard deviations for models 199

at most sites. Considering that the observed seasonal cycle is significantly different 200

from the model mean seasonal cycles for some sites (ULB, HFM, TGC, SAN) at 1-3 201

km height, synoptic variations (2-10 day) are consistent between models (correlation 202

coefficients generally greater than 0.5; Table S1). While the behaviour of seasonal 203

cycles depends on zonal mean emission distributions and seasonality in OH radical, the 204

synoptic variations in CH4 are produced by regional patterns of emissions and model 205

transport variability at the synoptic timescale. 206

207

The peak-to-trough seasonal cycle amplitude at 1-3 km and 3-5 km height ranges for 208

TGC, SGP and SUR are relatively large compared with the other sites (Table 3 and 209

Table S1). Greater daily CH4 variations during summer are observed at SUR, PFA, LEF 210

and ULB near the surface, coinciding with the seasonal maximum of CH4, and are 211

caused by the increase in surface emissions in summer and the seasonally-varying 212

dynamics associated with the planetary boundary layer height as well as cumulus 213

convection. On the other hand, troughs of the multi-model mean seasonal cycles at 214

TGC, HAA and the other remote sites sampling background air show lower variability 215

during the boreal summer months, coinciding with the seasonal minimum of CH4 in the 216

troposphere. In general, the seasonal amplitudes in the lower troposphere are larger than 217

those in the upper troposphere. The seasonal amplitude at the northern central Pacific 218

site HAA (Figure 2e) decreases from 50 to 20 ppb with altitude, whereas the seasonal 219

amplitude at the southern central Pacific site RTA (Table S1) is almost 20 ppb at every 220

height. Throughout the northern hemisphere, the amplitude weakens upward as the 221

signal from the lower troposphere (LT) propagates to the middle/upper troposphere 222



9

(MT/UT). This weakening is especially remarkable at the tropical site TGC (Figure 2f), 223

where the CH4 seasonal cycle amplitude at 6-8 km range is reduced to 11% of that at 224

0-2 km (all model average of 79 ppb). On average, the amplitude decreased only 225

marginally at ULB, while at other sites the amplitude decreased between 30-60% at 6-8 226

km compared to that at 0-2 km. Note here that the anthropogenic or biogenic emissions 227

around the Mongolian site ULB is among the lowest observed over the land regions. 228

229

One unusual feature among the seasonal cycles is observed as an increase in CH4 at 6-8 230

km height over SUR (73oE, 61oN) during May-Jun-Jul. The timing of this peak is about 231

two months ahead of the seasonal peak seen near the surface and there is almost no 232

seasonal variation in the 3-5 km altitude. An examination of CH4 longitude-pressure 233

cross sections (latitude range: 60-65oN; Figure S2) suggests that the high CH4 over SUR 234

during the late spring/early summer is transported from Western Europe. Air masses 235

carrying the signature of European emissions are lifted over the Ural mountain range 236

(57oE, 51oN to 66oE, 68oN), and are subsequently transported to Siberia, Mongolia and 237

as far as western North America by the prevalent westerly winds, as suggested by the 238

longitude-pressure cross-sections of CH4 concentrations (Figure S2). 239

240

3.2. Vertical profiles 241

Figure 3 shows the seasonal-mean vertical CH4 profiles at SUR, HAA, TGC and SAN 242

with the CTL fluxes during boreal winter (January-March: JFM) and summer 243

(July-September: JAS). The modeled (color lines) and observed (black dots with 244

horizontal bars) vertical profiles agree well over areas dominated by background 245

concentrations. Most model profiles at HAA and TGC fall inside the standard deviation 246

(black bars) of the seasonal variation of the observed CH4 concentrations. However, the 247

simulated vertical profiles at SAN show larger differences between models (a spread of 248

50 ppb at all altitudes) and with measurements. Due to sparse measurement at SAN, the 249



10

significance of model-observation differences cannot be assessed here. CH4 250

concentrations near the surface are affected by biospheric or anthropogenic activities as 251

well as meteorological conditions. For example, large synoptic variability is observed 252

near the surface at TGC during JAS months as the site is either under the influence of 253

marine air from the Atlantic Ocean from the Southeast (seen as relatively low 254

concentrations) or from regional emissions hotspots around the Gulf of Mexico. During 255

the JFM months, winds are predominantly southwesterly and the TGC site is under the 256

influence of emissions in Mexico and Southwestern USA. Thus the concentrations are 257

relatively stable (lower variability) and averaged higher than those in the JAS months. 258

At SUR, except below 2 km, the simulated results are generally lower than the observed 259

CH4 concentrations during both seasons. A more contrasting feature is observed in the 260

JAS season when the CH4 concentration is higher in the altitude range of 5.5-8 km 261

(upper troposphere) compared to those at 2-4 km (middle troposphere). Such elevated 262

levels of gases, with emissions only on the Earth’s surface, occur frequently in the 263

tropical upper troposphere in the presence of deep cumulus convection [e.g., Patra et al., 264

2009], and also seen in Fig. S2 for this high latitude region in Siberia. Only ACTM 265

simulation captures this feature to some extent. This is a clear indication that 11 out of 266

12 models underestimate vertical transport of surface emission signal to the upper 267

troposphere over Surgut due to deep cumulus convection. 268

269

The concentration at SAN sharply decreases with altitude in the lower troposphere 270

suggesting significant CH4 emission at this location, and the profiles above 3 km are 271

relatively straight in both seasons indicating strong vertical motion in all seasons over 272

this tropical site (Fig. 3g,h). Unfortunately, the data coverage is limited to only a few 273

profiles each season. When convective activity is intense during the boreal summer, 274

relatively straight vertical profiles are also observed at HAA, TGC and other NH sites. 275

The profiles at ocean sites HAA and RTA usually reveal characteristics of a background 276



11

site with insignificant surface emissions around the site during all seasons. However, 277

emission signals from Equatorial Asia/Western Pacific and Temperate Asia reach HAA 278

during October-January and April-July, respectively, at 850 hPa height (Fig. S2). In 279

addition, the Hawaiian Islands are located at the sub-tropical mixing barrier for 280

chemical species [Miyazaki et al., 2008], causing greater seasonal variability at HAA 281

than RTA (Table 3). Thus, the vertical profiles show a gradient in the JFM months 282

despite being an ocean site due to the influence of northern hemispheric emissions and 283

weak OH loss (Figure 3e). The CH4 peak at 3-4 km during the JAS months over HAA 284

(Fig. 4f), typical of all years, is caused by the transport of high CH4 from Boreal Asia 285

along the isentropic surfaces when the convective activity is low. Four models 286

(GEOS-Chem, PCTM, IFS and TM5) are able to simulate this feature to some extent. 287

GEOS-Chem and PCTM are driven by the meteorological reanalysis from the Goddard 288

Earth Observing System Data Assimilation System, where as IFS and TM5 use 289

ERA-Interim meteorology from the European Centre for Medium-Range Weather 290

Forecasts. 291

292

3.3. Differences between the observations and model simulations 293

Figure 4a-d show the model-observation scatter plots at 4 different height ranges (black: 294

0.5−1.5 km, blue: 1.5−3.0 km, green: 3.0−5.0 km and red: 5.0−7.0 km) at sites SUR, 295

THD, SGP and HAA. The black diagonal line intersects at the coordinate origin with 296

slope 1. Data points for the simulated concentrations are sampled from the nearest time 297

and altitude to the location of the measurements. The nearest altitude is linearly 298

interpolated between heights above and below the in situ sampling height. Correlation 299

coefficients R between the model and observation time series are presented in Table S1. 300

Several values, denoted by ‘0’, did not satisfy a significance level of 0.05 based on the 301

critical value for testing the null hypothesis H0 of non-correlation between two time 302

series. From the correlation coefficients, the variance of the model-observation 303



12

difference is larger near the surface than in the free troposphere at all sites due to large 304

signals at a daily timescale. 305

306

The model-observation difference at HAA in Figure 4d (also CAR, RTA and ULB) falls 307

within a small range of CH4 concentrations (±15 ppb, 1σ), and has a higher correlation 308

(generally greater than 0.7 in Table S1, except NIES08i and TOMCAT) than at the 309

other sites because of their location in the background marine environment (RTA and 310

HAA) and inland (CAR and ULB) away from direct influence of continental sources 311

(Figure 1). The model-observation differences at SUR, SGP and THD (also SAN, LEF, 312

HFM and PFA) are as large as 100 ppb for the high CH4 concentrations near the Earth’s 313

surface (Fig. 4). 314

315

Figure 4e-h shows the frequency distribution of the model-observation difference at 316

different seasons in winter (black: JFM), spring (blue: AMJ), summer (green: JAS) and 317

autumn (red: OND) for the analysis years of 2001-2007. The frequency distribution for 318

the small model-observation mismatches in the regions around HAA (Figure 4e), CAR, 319

RTA and ULB (not shown) shows little or no skewness and indicates little 320

seasonally-varying model biases. The skewness at THD is also small and the kurtosis is 321

low compared with Gaussian. On the other hand, SGP shows negative skewness, and 322

the peaks of the population densities in summer (blue and green histograms in panel g) 323

are biased negative by about 20 ppb. Thus Figure 4c and g confirm that models 324

simulated relatively low CH4 concentrations compared to the observations at SGP. At 325

SUR, the peak in summer (green in panel h) is marginally skewed positive due to an 326

overestimation of the modeled CH4 concentrations. These results suggest that the CH4 327

prior emissions are too low (marginally high) in the CTL model simulations around 328

SGP (SUR) site. These biases are unlikely to be due to transport model errors, because 329

most models systematically show similar under/over-estimations, and performance of 330



13

the TransCom-CH4 model ensemble has been found to be satisfactory at all other sites. 331

In addition, the lifetime of atmospheric CH4 is longer than 1 year at all layers in the 332

troposphere, thus the model biases between sites within a hemisphere/continent are not 333

likely to arise from uncertainty in OH abundance. However, it has been shown that the 334

latitudinal distribution of OH plays critical role in determining CH4 interhemispheric 335

gradient [Patra et al., 2011]. 336

337

3.4. Vertical, latitudinal and longitudinal gradients 338

We show the annual-mean modeled and observed vertical gradients in Figure 5 and 339

Table S2. Here, the vertical gradient in CH4 profiles is defined as the difference of 340

concentrations between the 1−3 km and 4−6 km heights. At oceanic (i.e. RTA and 341

HAA), coastal (i.e. TGC, THD, ESP and PFA) and continental (i.e. CAR and ULB) 342

sites, the simulated vertical gradients are generally within 5 ppb of the observed 343

gradients for most models. However, at SAN, SGP and HFM the differences between 344

simulated and observed vertical gradients are more than 5 ppb. At SUR and SGP, the 345

model-observation differences are systematically biased positive and negative, 346

respectively, for all the transport models. These results are consistent with over and 347

under estimation of simulated concentrations compared to observations in the lower 348

troposphere at SUR and SGP, respectively (Fig. 4c-h). These systematic 349

model-observation biases are thus likely to be caused by regional emissions in the CTL 350

and EXTRA scenarios rather than errors in CH4 loss and model transport. 351

352

Figure 6 shows comparisons of the simulated and observed latitudinal distributions of 353

CH4 concentrations among all sites at different height ranges (1.0−2.5 km, 2.5−5.0 km 354

and 5.0−8.0 km). The measured CH4 concentrations increase from SH to NH between 355

20oS to 50oN in the troposphere, at near linear rates of 1.4 ppb/degree at 1.0−2.5 km, 1.3 356

ppb/degree at 2.5−5.0 km and 1.2 ppb/degree at 5.0−8.0 km. These values are generally 357



14

well simulated by the models. The differences in latitudinal gradient are mainly due to 358

the decrease in CH4 concentrations with height in the NH, e.g., the upper tropospheric 359

concentration in the NH high latitude sites is ~30 ppb lower than the lower tropospheric 360

values. The agreement between the fitted straight line and the observations is best 361

between 2.5−5.0 km, where the standard deviation between these quantities amounts to 362

5 ppb (see also Table S3). Although model uncertainty in the free troposphere is usually 363

small, variance among the models at ULB at 5-8 km is larger than that in the other sites. 364

This is due to the long-range transport of European CH4 emission signal to ULB as 365

discussed earlier. The smaller (1.2 ppb/degree) interhemispheric gradient at 5-8 km, 366

compared to that at 2.5-5.0 km heights, indicates that the CH4 in the NH upper 367

troposphere is more readily transported to the SH. This effect can be seen as the higher 368

CH4 concentration in the upper troposphere at RTA than in the lower troposphere. 369

Higher concentration in the SH upper troposphere compared the lower troposphere is 370

also observed for nonreactive species, like carbon dioxide (CO2) and sulphur 371

hexafluoride (SF6) [Nakazawa et al., 1991; Miyazaki et al., 2008; Patra et al., 2009]. 372

The role of surface emissions on concentrations at sites is weakened rapidly with 373

increasing height between 1.0-2.5 km and 2.5-5.0 km (Fig. 6b,c). 374

375

Figure 7 shows the variation of CH4 with longitude for sites north of 42°N (i.e., SUR, 376

ULB, PFA, ESP, LEF and HFM) in the same height ranges as Figure 6 in order to 377

investigate zonal CH4 transport and vertical propagation driven by surface CH4 378

emissions from North America, Europe and Siberia. The lower tropospheric variations 379

in CH4 are mainly linked to stronger emissions around SUR and HFM compared to 380

those around ULB and ESP. The CH4 concentrations in the middle and upper 381

troposphere clearly have an eastward decreasing trend in observation and models (about 382

−0.1 ppb/degree). The high CH4 concentration in the free troposphere over Siberia is 383

diluted through mixing and reduces the difference between SUR and ULB at 5-8 km 384



15

height range. Apparently the CH4 rich air over the Eurasian continent in the upper 385

troposphere extends to the western side of the North American continent (PFA and 386

ESP) as suggested by the longitude-pressure cross section plots (ref. Fig. S1). The air in 387

the upper troposphere on the eastern side of Northern USA mostly originates at about 388

40oN over the Pacific Ocean, except for the winter months, giving relatively low CH4 389

concentrations at HFM. These results are generally consistent with Xiong et al. [2010], 390

but our understanding is improved significantly on the longitudinal variations in upper 391

tropospheric CH4 by including a greater number of vertical profile measurements. 392

393

3.5. Correlation of latitudinal and vertical transports 394

The over or under-estimations of latitudinal gradients of long lived atmospheric species 395

simulated by models, compared to the measurements, are thought to be linked with the 396

vertical concentration gradients [Denning et al., 1999] of those species. To explore this 397

hypothesis further, we show scatter-plots of vertical gradient at each site and latitudinal 398

gradient, with reference to HAA, within two altitude ranges 3-5 km and 0.5-2.5 km 399

(Fig. 8). Both gradients are plotted as the differences relative to the measured gradients, 400

showing which models over or under-estimate the vertical and latitudinal gradients. 401

Relative gradients are chosen to minimize effect of the model biases with respect to 402

measurement. In the middle troposphere (3-5 km), relative vertical and latitudinal 403

gradients are scattered in the approximate ranges of −10 to +7 ppb/km and −15 to +20 404

ppb/degree, respectively. Loose correlation between vertical and horizontal gradients 405

suggests that the interhemispheric transport is relatively insensitive to the local/regional 406

vertical profile of CH4 in the 3-5 km altitude range and that the transport in the models 407

is similar. However, the relative latitudinal gradients are apparently proportional to the 408

relative vertical gradient in the lower troposphere (0.5-2.5 km), suggesting that the 409

vertical transport of emissions through the lower troposphere plays a critical role in 410

determining interhemispheric gradients near the Earth’s surface. 411



16

412

A simpler representation of the relationship between vertical and interhemispheric 413

transport is shown in Figure 9a. The monthly-mean modeled vertical gradients in the 414

NH are calculated as the difference between the CH4 concentrations between 900 hPa 415

and 500 hPa altitudes. Values are also averaged over all the model grid cells between 416

30°N and 50°N for 2007. Average latitudinal gradients are calculated as the difference 417

between mean CH4 for SH (50°S −30°S) and NH (30°N −50°N). Annual mean values of 418

interhemispheric and NH vertical gradients in CH4 show a compact relationship (black 419

symbols). The seasonal variations in this relationship are mainly controlled by the 420

vertical transports, the emission and loss rates of CH4. In July, both the latitudinal and 421

NH vertical gradients are smaller than those in the other seasons because of faster loss 422

rates in the NH and lower loss rates in the SH as well as stronger vertical mixing in the 423

NH, despite high emission rates in the NH. In October, the NH vertical gradients are the 424

largest among all seasons (high emission rate, low loss scenario), although the 425

latitudinal gradients do not increase significantly. In January, the mid-tropospheric CH4 426

concentrations reach a seasonal high because of accumulation in the NH through the 427

autumn and early winter (low emission, very low loss scenario and weaker vertical 428

mixing). This condition led to the largest interhemispheric gradients in January. 429

430

The interactions between vertical and latitudinal gradients in CH4 (also valid for tracers 431

in the troposphere with lifetimes greater than the interhemispheric transport time of 432

about 1.3 years, see supplement for SF6 results in Figure S4) are schematically depicted 433

in Fig. 9b. Based on our analysis, we can hypothesize that (1) vertical transport is driven 434

by the strength of convection as well as turbulent mixing in the boundary layer BL), and 435

(2) meridional gradients depend mainly on the mixing in the BL for sites close to 436

emissions and on the horizontal advection scheme for the remote/background sites. 437

However, there is interaction between the turbulent mixing, convection and advection 438



17

processes in the model. A study using ACTM [Patra et al., 2009b] suggested that the 439

turbulent mixing transport emission signals to the 900-800 mb height layer, and the 440

convective transport species from the BL to the middle and upper troposphere. Most of 441

the interhemispheric transport due to advection occurs at higher altitudes. Thus a model 442

transport with weaker turbulent mixing and convection has a tendency towards steeper 443

meridional gradient in the simulated concentrations at lower altitudes but not at higher 444

altitudes (6 – 8 km). 445

446

As discussed in Patra et al. [2011] and seen in Fig. 9a, the TM5 model produces one of 447

the largest latitudinal gradients in CH4 among all the models that participated in the 448

TransCom-CH4 experiment. However, we find in this study that the NH vertical 449

gradients simulated by TM5 are not the largest. The latitudinal gradients shown in 450

Figure 6 suggest that the stronger gradients in the TM5 simulations are mainly due to 451

low SH concentrations compared to other models. The PCTM produced the largest 452

vertical gradients in the NH, although the simulated inter-hemispheric gradients are not 453

extreme. Stronger vertical gradients in both the hemispheres result in an 454

inter-hemispheric gradient that is close to the multi-model average. All other models 455

suggest that a stronger vertical gradient in NH troposphere corresponds to a stronger 456

inter-hemispheric gradient in CH4 concentrations. 457

458

459

4. Conclusions 460

We assessed latitudinal and vertical transports of tropospheric CH4 using the chemical 461

transport models (CTMs) compared with airborne in situ measurements. The 462

model-model difference is large near the surface in source regions, and small in the 463

middle/upper troposphere and over the marine background regions. The 464

model-observation differences are controlled predominantly by the model transport and 465



18

the surface emission rather than the loss rates of CH4. At most sites the simulated 466

concentrations vary around the observed values, while at two sites the model results are 467

systematically biased relative to the observed value. The positive and negative biases, 468

systematic for all models, can be attributed to higher and lower CH4 emissions than 469

found in the case of Surgut in Russia and South Great Plain in USA, respectively. These 470

results are consistent for the use of two emission scenarios, CTL or EXTRA. 471

472

Observed CH4 profiles show near zero vertical gradients above 3 km where models 473

indicate convective activity is the primary driver of mixing (all seasons over the 474

Amazon and summer months in the higher latitudes), and elevated concentrations in the 475

upper troposphere region, when measurements are available over Surgut. In contrast, 476

stronger decreases with altitudes above 3 km are observed in the mid- and high latitude 477

land regions where the anthropogenic emissions are intense (e.g., TGC, HFM, SUR) 478

compared to the lower emissions over the background marine sites (e.g., HAA, ULB). 479

The model-model differences as well as the mismatches with observations are found to 480

be greatest at sites located over intense emission regions for the height below 3 km. In 481

the middle and upper troposphere (MUT), greater model-model differences are found 482

over the regions with stronger convective activities (e.g., SAN, and NH sites during the 483

summer season). These two results suggest that the turbulent mixing (acting below 3 484

km) and deep cumulus convection (acting in the MUT region) parameterizations in the 485

CTMs are less similar compared to the other transport processes, such as advection. 486

487

Interhemispheric gradients of the simulations and observations were 1.5±0.1 and 488

1.3±0.2 ppb/degree for the lower and middle troposphere, respectively. In an effort to 489

understand the origin of model-model differences in interhemispheric gradients, we 490

have shown scatter plots of vertical gradients in CH4 concentrations between lower and 491

middle troposphere and latitudinal gradients between the two hemispheres. Generally, 492



19

the models that produced higher vertical gradients for CH4 (also SF6) concentrations 493

also produced steeper interhemispheric gradients near the surface. In models with 494

underestimated vertical mixing mainly due to boundary layer turbulent mixing and 495

parameterized convection, tracers tend to get trapped near the sources. In this case, the 496

model is likely to overestimate meridional gradients in tracer concentrations, resulting 497

from relative inefficient advective transport between the hemispheres. 498

499

500

Acknowledgments. 501

This work is supported by JSPS/MEXT KAKENHI-A (grant number 22241008). 502

Annemarie Fraser is supported by the UK Natural Environment Research Council 503

National Centre for Earth Observation. We acknowledge the work of John McGregor 504

and Marcus Thatcher in the development of CCAM. CCAM simulations were 505

undertaken as part of the Australian Climate Change Science Program and used the NCI 506

National Facility in Canberra, Australia. Ronald Prinn and Matthew Rigby are 507

supported by NASA-AGAGE Grants NNX07AE89G and NNX11AF17G to MIT. 508

Matthew Rigby is also supported by a NERC Advanced Fellowship. The TOMCAT 509

work at University of Leeds was supported by NERC/NCEO. The research leading to 510

the IFS results has received funding from the European Community's Seventh 511

Framework Programme (FP7 THEME [SPA.2011.1.5-02]) under grant agreement 512

n.283576 in the context of the MACC-II project (Monitoring Atmospheric Composition 513

and Climate – Interim Implementation). 514

515



20

516

References 517

Belikov, D., S. Maksyutov, T. Miyasaka, T. Saeki, R. Zhuravlev, and B. Kiryushov 518

(2011), Mass-conserving tracer transport modelling on a reduced 519

latitude-longitude grid with NIES-TM, Geosci. Model Dev., 4, 207-222, 520

doi:10.5194/gmd-4-207-2011. 521

Belikov, D., et al. (2012), Off-line algorithm for calculation of vertical tracer transport 522

in the troposphere due to deep convection, Atmos. Chem. Phys. Discuss., 12, 523

20239-20289. 524

Corbin, K. D., and R. M. Law (2011), Extending atmospheric CO2 and tracer 525

capabilities in ACCESS, CAWCR Technical Report No. 35, The Centre for 526

Australian Weather and Climate Research, ISBN: 978-1-921826-177, Aspendale. 527

Denning, A. S., M. Holzer, K. R. Gurney, M. Heimann, R. M. Law, P. J. Rayner, I. Y. 528

Fung, S. Fan, S. Taguchi, P. Friedlingstein, Y. Balkanski, J. Taylor, M. Maiss, and 529

I. Levin (1999), Three-dimensional transport and concentration of SF6: A model 530

intercomparison study (TransCom-2), Tellus, 51B, 266–297, 1999. 531

Dlugokencky, E. J., R. C. Myers, P. M. Lang, K. A. Masarie, A. M. Crotwell, K. W. 532

Thoning, B. D. Hall, J. W. Elkins, and L. P. Steele (2005), Conversion of NOAA 533

atmospheric dry air CH4 mole fractions to a gravimetrically prepared standard 534

scale, J. Geophys. Res., 110, D18306, doi:10.1029/2005JD006035. 535

Chipperfield, M. P. (2006), New Version of the TOMCAT/SLIMCAT Off-Line 536

Chemical Transport Model: Intercomparison of Stratospheric Tracer Experiments, 537

Q. J. R. Meteorol. Soc., 132, 1179-1203, doi:10.1256/qj.05.51. 538

Emmons, L. K., et al. (2010), Description and evaluation of the Model for Ozone and 539

Related chemical Tracers, version 4 (MOZART-4), Geosci. Model Dev., 3, 43-67, 540

doi:10.5194/gmd-3-43-2010. 541



21

Fraser, A., C. Chan Miller, P. I. Palmer, N. M. Deutscher, N. B. Jones, and D. W. T. 542

Griffith (2011), The Australian methane budget: interpreting surface and 543

train-borne measurements using a chemistry transport model, J. Geophys. Res., 544

Geophys. Res., 116, D20306, doi:10.1029/2011JD015964. 545

Fung, I., J. John, J. Lerner, E. Matthews, M. Prather, L. P. Steele, P. Fraser (1991), 546

Three-dimensional model synthesis of the global methane cycle, J. Geophys. Res., 547

96(D7), 13,033–13,065. 548

Gatti, L. V., J. B. Miller, M. T. S. D’Amelio, A. Martinewski, L. S. Basso, M. E. Gloor, 549

S. Wofsy, and P. Tans (2010), Vertical profiles of CO2 above eastern Amazonia 550

suggest a net carbon flux to the atmosphere and balanced biosphere between 2000 551

and 2009, Tellus, 62B, 581–594. 552

Gent, P. R., S. G. Yeager, R. B. Neale, S. Levis, and D. A. Bailey (2009), 553

Improvements in a half degree atmosphere/land version of the CCSM, Clim. 554

Dynam., 79, 25–58, doi:10.1007/s00382-009-0614-8. 555

Hourdin, F., I. Musat, S. Bony, P. Braconnot, F. Codron, J.-L. Dufresne, L. Fairhead, 556

M.-A. Filiberti, P. Friedlingstein, J.-Y. Grandpeix, G. Krinner, P. Levan, and F. 557

Lott (2006), The LMDZ4 general circulation model: climate performance and 558

sensitivity to parametrized physics with emphasis on tropical convection, Clim. 559

Dynam., 27, 787–813, doi:10.1007/s00382-006-0158-0. 560

Ito, A., and M. Inatomi (2012), Use and uncertainty evaluation of a process-based 561

model for assessing the methane budget of global terrestrial ecosystems, 562

Biogeosciences, 9, 759–773, doi:10.5194/bg-9-759-2012. 563

Kawa, S. R., D. J. Erickson III., S. Pawson, and Z. Zhu (2004), Global CO2 transport 564

simulations using meteorological data from the NASA data assimilation system, J. 565

Geophys. Res., 109, D18312, doi:10.1029/2004JD004554. 566

Krol, M. C., S. Houweling, B. Bregman, M. van den Broek, A. Segers, P. 567

vanVelthoven, W. Peters, F. J. Dentener, and P. Bergamaschi (2005), The 568



22

two-way nested global chemistry-transport zoom model TM5: Algorithm and 569

applications, Atmos. Chem. Phys., 5, 417– 432. 570

Law, R. M., E. A. Kowalczyk, and Y.-P. Wang (2006), Using atmospheric CO2 data to 571

assess a simplified carbon-climate simulation for the 20th century, Tellus, Ser. B, 572

53, 427–437, doi:10.1111/j.1600-0889.2006.00198.x. 573

Machida, T., et al. (2001), Temporal and spatial variations of atmospheric CO2 mixing 574

ratio over Siberia, paper presented at Sixth International CO2 Conference, Sendai, 575

Japan. 576

Miyazaki, K., P. K. Patra, M. Takigawa, T. Iwasaki, and T. Nakazawa (2008), 577

Global-scale transport of carbon dioxide in the troposphere, J. Geophys. Res., 113, 578

D15301. 579

Nakazawa, T., K. Miyashita, S. Aoki, and M. Tanaka (1991), Temporal and spatial 580

variations of upper tropospheric and lower stratospheric carbon dioxide, Tellus B, 581

43, 106-117. 582

Olivier, J. G. J., and J. J. M. Berdowski (2001), Global emissions sources and sinks, in 583

The Climate System, edited by J. Berdowski, R. Guicherit, and B. J. Heij, A.A. 584

Balkema Publishers/Swets & Zeitlinger Pub., Lisse, The Netherlands, ISBN 585

9058092550, pp. 33-78. 586

Patra, P. K., et al. (2009a), Growth rate, seasonal, synoptic, diurnal variations and 587

budget of methane in lower atmosphere, J. Meteorol. Soc. Jpn., 87(4), 635-663. 588

Patra, P. K., M. Takigawa, G. S. Dutton, K. Uhse, K. Ishijima, B. R. Lintner, K. 589

Miyazaki, and J. W. Elkins (2009b), Transport mechanisms for synoptic, seasonal 590

and interannual SF6 variations and “age” of air in the troposphere, Atmos. Chem. 591

Phys., 9, 1209-1225. 592

Patra, P. K., S. Houweling, M. Krol, P. Bousquet, et al. (2011), TransCom model 593

simulations of CH4 and related species: linking transport, surface flux and 594



23

chemical loss with CH4 variability in the troposphere and lower stratosphere, 595

Atmos. Chem. Phys., 11, 12813-12837, doi:10.5194/acp-11-12813-2011. 596

Rotman, D. A., et al. (2004), IMPACT, the LLNL 3-D global atmospheric chemical 597

transport model for the combined troposphere and stratosphere: Model description 598

and analysis of ozone and other trace gases, J. Geophys. Res., 109, D04303, 599

doi:10.1029/2002JD003155. 600

Spivakovsky, C., et al. (2000), Three-dimensional climatological distribution of 601

tropospheric OH: Update and evaluation, J. Geophys. Res., 105(D7), 8931-8980. 602

Tans, P. P., et al. (1996), Carbon Cycle (Group Report), Climate Monitoring and 603

Diagnostics Laboratory, No. 23, Summary Report 1994-1995, Hoffman, D.J., J.T. 604

Peterson and R.M. Rosson, eds, US Department of Commerce, Boulder, 605

Colorado. 606

van der Werf, G. R., J. T. Randerson, L. Giglio, G. J. Collatz, P. S. Kasibhatla, and A. 607

F. Arellano Jr (2006), Interannual variability in global biomass burning emissions 608

from 1997 to 2004, Atmos. Chem. Phys., 6, 3423-3441, 2006. 609

Xiong, X., C. D. Barnet, Q. Zhuang, T. Machida, C. Sweeney, and P. K. Patra (2010), 610

Mid-upper tropospheric methane in the high northern hemisphere: Spaceborne 611

observations by AIRS, aircraft measurements, and model simulations. J. Geophys. 612

Res., 115, D19309. 613

Zhou, L.X., D. Kitzis, and P. P. Tans (2009), Report of the fourth WMO round-robin 614

reference gas intercomparison, 2002–2007. In: Report of the 14th WMO meeting 615

of Experts on Carbon Dioxide Concentration and Related Tracer Measurement 616

Techniques (ed. T. Laurila). Helsinki, Finland, September 10–13, 617

2007,WMO/GAWReport No. 186, 40–43. 618

619

620



24

Table 1. List of the CTMs used in this study (see text for further details), and basic 621

model features. 622

623

Model Name Contributing

Institute

Vertical

levels

Horizontal

resolution Source of Meteorology

ACTM RIGC 67σ ~2.8 × 2.8° NCEP2; U, V, T; SST

CAM CU 28σ 2.5 × ~1.9° NCEP/NCAR

CCAM CSIRO 18σ ~220 km NCEP; U, V; SST

GEOS-Chem UoE 30/47η 2.5 × 2.0° NASA/GSFC/GEOS4/5

IFS ECMWF 60η 0.7× 0.7° ECMWF, ERA-interim

IMPACT_1x1.25 LLNL 55η 1.25 × 1.0° NASA/GSFC/GEOS4

LMDZ LSCE 19η 3.75 × 2.5° ECMWF; U, V, T; SST

MOZART MIT 28σ ~1.8 × 1.8° NCEP/NCAR

NIES08i NIES 32σ-θ 2.5 × 2.5° JCDAS, ERA-interim-PBL

PCTM GSFC 58η 1.25 × 1.0° NASA/GSFC/GEOS5

TM5_1x1 SRON 25η 1.0 × 1.0° ECMWF, ERA-interim

TOMCAT UoL 60η ~2.8 × 2.8° ECMWF,ERA-40/interim

624

625



25

Table 2. Site details of the airborne in situ measurements used in this study. All sites, 626

except for Surgut, are operated by the NOAA/ESRL. 627

628

Site Location Latitude Longitude AMSL (m) Period

PFA Poker Flat 65.1°N 147.3°W 1200−7400 2001−2007

SUR Surgut 61.0°N 73.0°E 500−7000 2001−2007

ESP Estevan Point 49.6°N 126.4°W 300−5500 2002−2007

ULB Ulaanbaatar 47.4°N 106.2°E 1500−6000 2004−2007

LEF Park Falls 45.9°N 90.3°W 600−4000 2001−2007

HFM Harvard Forest 42.5°N 72.2°W 600−7700 2001−2007

THD Trinidad Head 41.1°N 124.2°W 300−7700 2003−2007

CAR Briggsdale 40.9°N 104.8°W 2100−8700 2001−2007

SGP Southern Great Plains 36.8°N 97.5°W 170−4900 2006−2007

TGC Sinton 27.7°N 96.9°W 300−7700 2003−2007

HAA Molokai Island 21.2°N 159.0°W 300−7700 2001−2007

SAN Santarem 2.9°S 55.0°W 150−5200 2001−2005

RTA Rarotonga 21.3°S 159.8°W 300−6100 2001−2007

629

630



26

Table 3. Measured and model simulated seasonal cycle amplitudes of CH4 at two 631

altitude ranges. Model results are given as mean and 1-σ standard deviations (individual 632

model results are available in Table S1). 633

634

Sites 1-3 km 3-5 km

Observed Model Observed Model

Mean SD Mean SD

PFA 28.9 31.6 7.5 28.9 24.6 4.7

SUR 56.4 49.9 16.9 39.5 18.6 5.1

ESP 34.9 40.1 4.5 33.4 24.2 6.9

ULB 35.0 20.5 6.7 35.2 16.2 3.3

LEF 32.4 27.5 6.5 32.9 30.2 5.1

HFM 48.9 26.8 7.0 34.6 28.8 6.6

THD 42.3 38.8 4.7 29.2 29.3 6.2

CAR -- -- -- 34.7 30.0 8.8

SGP 73.9 52.1 14.7 38.7 32.3 5.7

TGC 89.0 65.1 13.7 32.9 26.1 7.0

HAA 36.9 44.0 8.6 30.2 34.8 4.5

SAN 86.7 46.9 10.0 68.6 33.3 7.2

RTA 26.4 28.3 12.9 35.1 25.2 6.8

635

636



27

637 638

Figure 1. Sampling locations for the in situ aircraft measurements. Background colors 639

on the map show intensity of annual mean CH4 emission for the CTL flux. 640

641



28

642

Figure 2. Seasonal variations of the observed (cyan dots with error bar) and simulated 643

(colour lines; see legends in the bottom-left panel; continuous at daily intervals) CH4 644

concentrations are shown for a comparison at selected sites for 2007. The error bars are 645

showing the 1σ standard deviation of all data within a selected height range. Smooth 646

black lines present fitted curves (3 harmonics) to the average seasonal cycle for all the 647

models. The results for each site are averaged over 3 different altitude ranges, 648

representing lower troposphere (0.5-2 km), middle troposphere (3-5 km) and upper 649

troposphere (6-8 km). 650



29

651

Figure 3. The observed (symbol) and simulated (lines) CH4 vertical profiles at SUR 652

(top row), HAA (2nd row from top), TGC (3rd row from top) and SAN (bottom row) 653

for January-March (left) and July-September (right) in 2007. Horizontal bars for the 654

monthly-mean measurements present a 1-σ standard deviation. Results for 2005 and 655

2007 at more number of sites are shown in Fig. S3. 656

TGC, 2007, JFM TGC, 2007, JAS

SAN, 2003, JFM

HAA, 2007, JFM HAA, 2007, JAS

SAN, 2003, JAS

0

1

2

3

4

5

6

7

8

9

10

0

1

2

3

4

5

6

7

8

9

0

1

2

3

4

5

6

7

8

9

10

10

1750 1800 1850 1900 1750 1800 1850 1900

SUR, 2007, JFM SUR, 2007, JAS

ACTMCAMCCAMIFSGEOS-ChemIMPACT_1x1.25LMDZ _ _MOZART _ _NIES-08i _ _PCTM _ _TM5_1x1 _ _TOMCAT _ _Obs.

a

b

c

d

e fAltit

ude

(km

)

CH4 (ppb)

0

1

2

3

4

5

6

7

8

9

10

g h



30

657

Figure 4. Correlations of observed and simulated CH4 concentrations for the analysis 658

period (2001−2007) for 4 selected sites (SUR, THD, SGP, HAA are organized from top 659

to bottom row, respectively). The symbols within each panels of the left column 660

correspond to different heights: 0.5−1.5 (black), 1.5−3.0 (blue), 3.0−5.0 (green) and 661

5.0−7.0 km (red). Note that the axis range varies for panels a-d. The right panels show 662

the frequency averaged over the all CTMs at different seasons of JFM (black), AMJ 663

(blue), JAS (green) and OND (red), where the class interval is 2 ppb. 664



31

665

Figure 5. Difference between the observed and simulated CH4 concentrations averaged 666

for the analysis years (2001−2007) are shown for the lower troposphere (a; 1.0 km to 667

3.0 km) and middle troposphere (b; 3.0 to 5.0 km). All values are offset by the yearly 668

average CH4 concentrations over 1-6 km at HAA before calculating the differences. 669

Results corresponding to CTL and EXTRA fluxes are shown in blue and red, 670

respectively. Bottom panel (c) shows the vertical gradients of the CH4 concentrations 671

between 1-3 km and 4-6 km altitude ranges. 672

ACTMCAMCCAMIFSGEOS-ChemIMPACT_1

LMDZMOZARTNIES-08iPCTMTM5_1TOMCATObs

RTA SANHAATGC SGP CAR THDHFM LEF ULB ESP SUR PFA

-5

0

5

10

15

20

25

Vert

ical

Gra

dien

t, pp

b/km

-40

-20

0

20

40

60

Di!

eren

ce, p

pbD

i!er

ence

, ppb

a (3-5 km)

b (1-3 km)

c

-40

-20

0

20

40

60



32

673

Figure 6. The CTMs-simulated CH4 concentrations averaged in the upper troposphere 674

of 5.0-8.0 km (a), the middle troposphere of 2.5-5.0 km (b) and the lower troposphere of 675

1.0-2.5 km (c) for the analysis years (2001−2007) with the CTL (blue) and EXTRA 676

(red) fluxes, and observed concentrations (black). 677

a

b

c

La tude

CH

4 (

pp

b)

ACTMCAMCCAMIFSGEOS-ChemIMPACT_1

LMDZMOZARTNIES-08iPCTMTM5_1TOMCATObs

SAN

RTA

HAA

TGC

SUR

PFA

SGP

ESP

CARTHD

HFMLEF

ULB

-30 -20 -10 0 10 20 30 40 50 60 701700

1725

1750

1775

1800

1825

1850

1875

19001700

1725

1750

1775

1800

1825

1850

1875

1900

1700

1725

1750

1775

1800

1825

1850

1875

1900



33

678

679

Figure 7. Londitudinal distribution of the CTMs-simulated CH4 concentrations at sites 680

higher than 42°N (i.e., SUR, ULB, PFA, ESP, LEF and HFM from 0°E to 360°E), 681

averaged in the upper troposphere of 5.0-8.0 km (a), the middle troposphere of 2.5-5.0 682

km (b) and the lower troposphere of 1.0-2.5 km (c) for the analysis years (2001−2007) 683

with the CTL (blue) and EXTRA (red) fluxes, and observed concentrations (black). The 684

black line presets a robust fitting line of the observations (the fitting coefficients of all 685

the models are also seen in Table 3). 686



34

687

688

Figure 8. CH4 vertical gradients (ppb/km) are plotted against the latitudinal gradients in 689

CH4 (ppb/degree) between the sites within two altitude ranges (a) 3.0-5.0 km and (b) 690

0.5-2.5 km. All data are averaged for the period 2001−2007 and the results for CTL flux 691

are shown. The symbol for each model is colored coded for the gradients between each 692

site and the reference site HAA. Both the gradients are plotted by a difference relative to 693

the measured gradients. 694

695



35

696

697

Figure 9. (a) Modeled vertical gradient in the NH between 900 hPa (~1 km) and 500 698

hPa (~5 km) in the latitude range of 30°N −50°N are plotted versus the modeled 699

latitudinal gradient between the NH (30°N −50°N) and the SH (50°S −30°S), averaged 700

over the height range of 900-500 hPa (~1.0−5.5 km). Results are shown for different 701

months, representing boreal winter, spring, summer and autumn (January, April, July 702

and October, respectively) and for annual mean. (b) A schematic of the relationship 703

between the vertical and interhemispheric transports is depicted in the lower panel. 704

transcom model simulations of methane: comparison of

Documents