We present an analysis of interannual variability (IAV) and trends in atmospheric methane (CH4) mixing ratios over the western Pacific between 55°N and 35°S from 1994 to 2010. Observations were made by the Center for Global Environmental Research (CGER) of the National Institute for Environmental Studies (NIES), using voluntary observation ships sailing between Japan and Australia/New Zealand and between Japan and North America, sampling background maritime air quasi-monthly (∼10 times per year) with high latitudinal resolution. In addition, simulations of CH4 were performed using NIES atmospheric transport model. A large CH4 increase was observed in the tropics (10°N–5°S) during 1997 (between 15 ± 3 and 19 ± 3 ppb yr−1) and during 1998 for other regions (40°N–50°N: 10 ± 2–16 ± 1 ppb yr−1; 10°S–25°S: 12 ± 2–22 ± 4 ppb yr−1). The CH4 increase leveled off from 1999 to 2006 at all latitudes. The CH4 growth rate was enhanced in 2007 (25°N–50°N: 10 ± 1–12 ± 3 ppb yr−1; 15°S–35°S: 7 ± 1–8 ± 1 ppb yr−1) but diminished thereafter; however, a large CH4 growth (10 ± 1–17 ± 1 ppb yr−1) was observed in 2009 over the northern tropics (0°–15°N). These observations, combined with the simulation results, suggest that to explain the CH4 increase in 2007 would require an increase in surface emissions of ∼20 ± 3 Tg-CH4 yr−1 globally and an increase in the Northern Hemisphere (NH) of 4–7 ± 3 Tg-CH4 yr−1 more than that in the Southern Hemisphere (SH), assuming no change in OH concentrations; alternatively, a decrease in OH concentrations of 4.5 ± 0.6%–5.5 ± 0.5% yr−1 globally would be required if we assume no change in surface emissions. Over the western Pacific, the IAV in CH4 within the northern tropics was characterized by a high growth rate in mid-1997 and a reduced growth in 2007. The present data indicate that these events were strongly influenced by the IAV in atmospheric circulation associated with El Niño and La Niña events. Our observations captured the CH4 anomaly in 1997 associated with forest fires in Indonesia. The IAV and trends in CH4 as seen by our data sets capture the global features of background CH4 levels in the northern midlatitudes and the SH, and regional features of CH4 variations in the western tropical Pacific.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Methane (CH4) mixing ratios in the atmosphere have shown a steady increase in recent decades [e.g., Forster et al., 2007]. The global annual average of the CH4 mole fraction has risen from 1645 parts per billion (ppb) in 1984 to 1773 ppb in 1999 [Dlugokencky et al., 2003, 2005]; however, it has remained almost constant since 1999 [e.g., Dlugokencky et al., 2003; Simpson et al., 2006]. The trends in CH4 mixing ratios are affected by the balance between emissions from anthropogenic sources (e.g., fossil fuels, ruminant animals, rice paddies, and biomass burning) and from natural sources (mainly wetlands, but also other natural sources such as termites) on the Earth's surface, and sinks, mainly losses via reactions with the hydroxyl radical (OH) in the atmosphere. Our understanding of these processes remains insufficient to explain the lack of growth in the CH4 mixing ratio during 1999–2006 [Forster et al., 2007].
 CH4 mixing ratios also show interannual variations. For example, anomalously high growth rates of ∼15 ppb yr−1 were observed globally in 1991 and 1998 [e.g., Dlugokencky et al., 1996, 2001; Langenfelds et al., 2002; Simpson et al., 2006]. The high CH4 growth rate in 1991 was attributed to reduced OH sinks resulting from the eruption of Mt. Pinatubo [Dlugokencky et al., 1996]. During 1997–1998, when a strong El Niño event was occurred, increased emissions from wetlands and the burning of boreal biomass are likely to have contributed to the anomalously high CH4 mixing ratios [Dlugokencky et al., 2001; Langenfelds et al., 2002; Simpson et al., 2006]. Bousquet et al.  suggested that interannual variability (IAV) in atmospheric CH4 is dominated by wetland emissions. These previous studies investigated the causes of IAV in atmospheric CH4 based on analyses of the IAV in CH4 sources and sinks. In contrast, Patra et al.  proposed that the IAV of atmospheric CH4 in tropical regions is due mainly to the IAV of atmospheric transport, based on the results of simulations using an atmospheric general circulation model-based chemical transport model (ACTM). Chen and Prinn  showed that the IAV of atmospheric transport during the El Niño event of 1997–1998 influenced seasonal cycles of CH4 at Cape Matatula, Samoa.
 Recent measurements have shown large growth in CH4 mixing ratios starting from around the beginning of 2007 [Rigby et al., 2008; Dlugokencky et al., 2009]. Rigby et al.  found that the growth occurred simultaneously, and with similar rates of ∼10 ppb yr−1, at all 13 ground stations of the Advanced Global Atmospheric Gases Experiment (AGAGE) and the Commonwealth Scientific and Industrial Research Organization (CSIRO) networks in both hemispheres. Their analysis was based on a 12-box model, and they concluded that an emission increase in both hemispheres was required to explain the increase in CH4 mixing ratio in 2007, assuming no change in the OH concentration. The authors also showed that the emission increase would be higher in the Northern Hemisphere (NH) than in the Southern Hemisphere (SH) in the case of a small decrease in the OH concentration. Dlugokencky et al.  showed that the globally averaged CH4 mole fraction increased by 8.3 ± 0.6 ppb in 2007 and by 4.4 ± 0.6 ppb in 2008, based on data from the measurement network of the National Oceanic and Atmospheric Administration's Global Monitoring Division (NOAA/GMD). They suggested that emissions from wetlands were enhanced by anomalously high temperatures in the northern high latitudes and by large precipitation in the tropics during 2007.
 The purpose of this study is to present an analysis of IAV and trends in the atmospheric CH4 mixing ratio over the western Pacific region between 55°N and 35°S from October 1994 to June 2010, as observed by the Center for Global Environmental Research (CGER) of the National Institute for Environmental Studies (NIES) based on data collected by voluntary observation ships (VOS). Air sampling was performed quasi-monthly (∼10 times per year) with high latitudinal resolution. There are few ground-based stations around the western tropical Pacific (e.g., Guam by NOAA/GMD; Cape Matatula by AGAGE and NOAA/GMD). The global monitoring network run by the University of California–Irvine has sampled the western tropical Pacific every 3 months since the 1980s [Simpson et al., 2002]. NOAA has performed shipboard measurements in the western Pacific since 2004, although the sampling frequency is just once or twice per year. Our observational data, collected over the western Pacific across the tropics with a higher spatiotemporal resolution than previous data sets, provide unique and important information on the distribution and variability of background CH4 mixing ratios in the maritime atmosphere.
 Here, we focus on the characteristics of the large CH4 increases recorded during 1997–1998 (in section 3.1) and 2007 (in section 3.2). We use the output from a three-dimensional atmospheric transport model driven by realistic meteorological data to examine the cause of the observed variations in CH4. The meteorological fields in the western Pacific, especially in the tropics, differed between the El Niño period in 1997–1998 and the La Niña period in 2007–2008. We discuss the effect of the meteorological field on the IAV of CH4 growth rate in the western tropical Pacific in section 3.3.
2.1. Air Sampling
 NIES/CGER performs atmosphere and ocean observations using VOS along two transects between Japan and Australia/New Zealand (north-south cruise in the western Pacific) and between Japan and North America (east-west cruise in the northern Pacific). Figure 1 shows the location of air samplings collected by the VOS program. Air samples along the north-south cruise have been collected onboard the following regular cargo ships: the M/S Hakuba Maru of Nippon Yusen Kaisha (NYK) Line (herein referred to as HK; preliminary cruise in 1991 and continuous cruises from June 1992 to January 1996; we used only quality-controlled data for the period after October 1994), the Southern Cross Maru of Mitsui O.S.K. Lines (MOL), Ltd. (herein SC; March 1996 to May 2001), and the MOL Golden Wattle (GW; May 2001 to May 2002) sailing between Japan and Australia. Samples were also collected aboard the MOL Glory (MG; September 2002 to February 2003), the Fujitrans World of Kagoshima Shipping Co. (FW; February 2003 to September 2005), and the Transfuture 5 of Toyofuji Shipping Co., Ltd. (TF5; November 2005 to the present) sailing between Japan and Australia/New Zealand (Figure 1c). By combining the data from these ships, the coverage is nearly continuous from October 1994 to the present for the region between 30°N and 25°S, extended to 38°S for the period after 2003.
 The cargo ships undertake the round trip cruises every 5–6 weeks (∼10 times per year). Each cruise takes about 2 weeks each way (between Japan and Australia/New Zealand), with about 3 days spent in port at each country. The average interval between air samplings was 43 days for each latitude band. In the tropics, the sampling interval was reduced to 23 days after 2003.
 For the northern Pacific, air samples have been collected onboard the following regular cargo ships: the M/S Skaugran of Seaboard International Shipping Co., Ltd. (SK; March 1995 to September 1999), the Alligator Hope of MOL (AH; November 1999 to May 2001), the Pyxis of Toyofuji Shipping Co., Ltd. (PX; November 2001 to the present), and the Skaubryn of Seaboard International Shipping Co., Ltd. (SB; April 2005 to July 2010), sailing between Japan and North America (Figure 1b). The PX cruise takes about 2 weeks each way between Japan and the West Coast of the United States/Canada and 4 weeks each way between Japan and the East Coast of the United States, via the Panama Canal. The SK, AH, and SB cruises run between Japan and the West Coast only. The ships spend about 3 days in port at each country. In this study, we use air samples collected west of 180°.
 The average sampling interval for northern Pacific cruises (60 days) is longer than that for the north-south cruise. Because of periodic changes in the shipping routes, air sampling has been discontinued for some latitude bands and for some periods in the northern Pacific. From mid-2001 to early 2005, no samples were collected north of 48°N. The sampling frequency was low during 1998–1999 for all latitudes, and during 1997, 1999 and after 2005 for 35°N–40°N. The longitudinal range of sampling locations for east-west cruises is larger than that for the north-south cruises (Figure 1a).
 Two automatic sampling systems, using 3.3 L stainless-steel flasks and 2.5 L Pyrex glass flasks, were installed on the cargo ships. The sampling system using stainless flasks has been installed since 1994 (red circles in Figures 1b–1c) and the collected air samples have been used for laboratory measurements of the mixing ratios of carbon dioxide (CO2), nitrous oxide (N2O), CH4, and CO2 isotopes. Since early 2000, the sampling trigger has been controlled by a Global Positioning System (GPS) to sample the air at fixed latitudes. For north-south cruises by FW and TF5, 21 samples were collected using stainless flasks at 38°S, 34°S, 30°S, 26°S, 22°S, 18°S, 15°S, 12°S, 9°S, 6°S, 3°S, 0°, 3°N, 6°N, 9°N, 12°N, 15°N, 18°N, 22°N, 26°N, and 30°N on the northbound cruise from New Zealand to Japan. Before 2000, the system was controlled by a timer (HK and SC voyages) or manually (MG voyages) to sample the air at intervals of approximately 3° in latitude, meaning that the sampling latitudes varied among the cruises. For the northern Pacific, 14 samples were collected between Japan and North America.
 The sampling system using glass flasks has been installed since the end of 2001 (blue circles in Figures 1b–1c), to measure the mixing ratios of CO2, N2O, CH4, and other trace gases, and the O2/N2 ratio [Tohjima et al., 2005]. Twenty-one samples were collected using glass flasks on the round-trip cruises of FW and TF5. The sampling latitudes differed from those using stainless flasks: 7 samples were collected on the southbound cruise from Japan to Australia (18°N, 13°N, 8°N, 3°N, 2°S, 7°S, and 12°S), no sampling was performed between Australia and New Zealand, and 14 samples were collected on the northbound cruise from New Zealand to Japan (37°S, 32°S, 27°S, 22°S, 17°S, 12°S, 7°S, 2°N, 3°N, 8°N, 13°N, 18°N, 23°N, and 28°N). After 2003, the temporal interval of air sampling between 18°N and 12°S was reduced to about 20 days (shorter than that for other latitudes) because of the use of additional glass flask sampling on the southbound cruise. For the northern Pacific, we collected 7 samples by PX and 14 samples by SB.
 Air intakes were placed at the bow (bosun store) on HK and SC (about 10 m above sea level (asl)); on the bow mast on GW and AH (18 m asl); at the top of the bridge on MG (28 m asl), FW (28 m asl), TF5 (28 m asl), and PX (26 m asl); on the roof of a laboratory on SK (30 m asl); and on the roof of a gyro room on SB (28 m asl). For each ship, the air intakes are located at the bow or midship, and the smokestacks are located near the stern. For example, the distance between the air intake and smokestack is 163 m on TF5 and 56 m on PX. Thus, the ship emissions are expected to have a minor effect on the air samples, except on MG, for which the smokestack is located near the bridge. The air samples would have been contaminated by ship emissions in the case of tail winds, but contaminated data were removed based on analyses of CO2 mixing ratios (see section 2.3).
 The air for sampling was drawn from the intake by a metal bellows pump (MB-151, Metal Bellows). The air was then passed through a cold trap (−45°C) and collected in the 3.3 L stainless flasks at 0.25 MPa (gauge pressure) after a 5 min purge, and in the 2.5 L glass flasks at 0.1 MPa after a 20 min purge. For details of the method employed in sampling by glass flasks, see Tohjima et al. . The sampling lines were similar for both systems.
2.2. CH4 Measurements
 The CH4 mole fraction in the collected air samples was measured at NIES by gas chromatography with a flame ionization detector (GC-FID; HP-5890, Agilent Technologies) [Machida et al., 2008]. Before analysis, the air was dried by passing through a glass trap cooled to −80°C. The analytical repeatability of CH4 measurements (the average value of the standard deviation for repeated measurements) was ∼2 ppb, for both glass and stainless flasks.
 The NIES CH4 scale was established based on eight standard gases in high-pressure cylinders ranging between 500 and 3000 ppb prepared by the gravimetric method in 1994 (NIES 94 CH4 scale). In this paper, CH4 mixing ratios are reported as dry air mole fractions (nmol mol−1 or ppb) traceable to the NIES 94 CH4 scale. We mainly use six cylinders from 1250 to 2500 ppb in calibrating the working standard gases to measure mixing ratios of tropospheric CH4. The NIES 94 CH4 scale was compared with a scale developed by Tohoku University (TU) in 2004–2005, as part of an intercomparison performed by the World Calibration Centre for Methane in Asia and the South-West Pacific (http://gaw.kishou.go.jp/wcc/ch4/comparison.html). Our CH4 values were higher than the TU scale by 1.6 and 2.2 ppb at 1810 and 1960 ppb, respectively. The 2002–2007 WMO Round-Robin intercomparison [Zhou et al., 2009] revealed that the NIES 94 scale was higher than the NOAA 04 scale by 3.5–4.6 ppb in the range between 1750 and 1840 ppb.
 No significant biases in CH4 mixing ratios were observed between sampling by stainless flask and by glass flask. The mean difference in CH4 mixing ratio (stainless minus glass) and its standard deviation were −0.1 ± 4.0 ppb for 168 coincident samples at 12°N, 3°N, 12°S, and 22°S from 2003 to 2008. Thus, we combined the data from both types of flask.
2.3. Data Analysis
 From October 1994 to June 2010, approximately 4240 air samples were collected from the north-south cruises in the western Pacific. We removed data for samples from stainless flasks during 1999–2000 (all 126 samples from 6 cruises; SC 29, 30, 33, 36, 37, and 38) and for samples from glass flasks during 2006–2007 (all 105 samples from 5 cruises; TF5 9–13), as they showed anomalous CO2 mixing ratios and latitudinal distributions due to a leak in the sampling line. We screened a latitudinal profile of each cruise and removed samples with a CO2 mixing ratio that differed by 3 ppm or more compared with neighboring samples, correcting for the latitudinal gradient of CO2.
Figure 2 shows an example of CH4 measurements during a cruise from New Zealand to Japan. The sample collected at 30°N (red cross) was excluded from analysis because the CO2 mixing ratio was 398.94 ppm, which is ∼10 ppm higher than the value for samples collected at 22°–28°N. Contaminated samples were often observed near the coasts of Japan, Australia, or New Zealand. Approximately 200 samples were removed according to this criterion. The rejected data are shown in Figure 1c (open circles). As a next step, samples with a normal CO2 mixing ratio but a high CH4 mixing ratio were removed by data fitting of CH4 time series.
 Approximately 1760 air samples were collected during the cruises in the northern Pacific. We removed all 42 samples from 3 cruises (AH 04, 07, and 11). We also applied filtering based on the CO2 mixing ratio (samples were excluded if the CO2 mixing ratio differed from that of adjacent samples by 5 ppm or more), resulting in the removal of about 90 samples. However, in some cases the data screening was ineffective in terms of removing contaminated samples because the CO2 mixing ratio over the northern Pacific commonly shows very large longitudinal variations. Any remaining contaminated samples were removed by time series analysis.
 Finally, we analyzed approximately 3800 (north-south cruises) and 1620 (northern Pacific cruises) samples (closed circles in Figures 1b–1c). For the time series analysis, we grouped the north-south cruise data into 5°-latitude bins (±2.5° for the target latitude) between 30°N and 35°S, and grouped the northern Pacific cruise data into 10°-latitude bins (40° ± 5°N and 50° ± 5°N). The wider bin for the northern Pacific cruises was chosen because of the existence of gaps in the sampling.
 We applied a digital filtering technique [Nakazawa et al., 1997] to derive a fitting curve for the CH4 time series, as well as long-term trends, IAV, and the seasonal cycle. A 36 month (3 year) low-pass filter was used to obtain the long-term trends. The growth rate was calculated as the derivative of the long-term trend with respect to time. The averaged seasonal cycle is expressed as the sum of the Fourier harmonics with periods of 12, 6, and 4 months. The fitting curve is represented by the sum of the long-term trends with periods longer than 36 months, shorter irregular variations with periods between 4 and 36 months, and the averaged seasonal cycle. The data outside ±3 standard deviations from the fitting curve were regarded as outliers and removed from the analysis. The fitting procedure was repeated until no outliers remained in the time series. In this paper, we focus on the long-term trend and on the growth rate.
 We estimated the errors in the calculated long-term trend and growth rate arising from different sampling frequency (σSMP) and uncertainties in GC measurements (σGC). To estimate σSMP, we used a bootstrap method in which a number of data set with equal size to the original data set were prepared by random resampling. Note that there were overlapped data in each bootstrap data set and only ∼63% of the original data were selected in average. We prepared 100 data sets of resampled time series for each latitude band and calculated the long-term trend and growth rate by the same method. To estimate σGC, we prepared 100 data sets for each latitude, in which the uncertainties (1σ = ±2 ppb) were randomly added to the CH4 mixing ratios; subsequently, the long-term trend and growth rate were calculated. The σSMP and σGC of the long-term trend and growth rate were determined as a standard deviation of the 100 values.
 We derived a total error (σTOT) as σTOT2 = σSMP2 + σGC2, although the σGC component (about 25% of σSMP) is negligible in σTOT. The observed long-term trend and growth rate are provided along with σTOT in the following text and in Figure 3. The σTOT values are large at 50°N during 2001–2004 (due to a low sampling frequency) and at 25°N and 30°N (due to large seasonal variations in CH4 mixing ratios). After 2003, the σTOT values become smaller at almost all latitudes because of the addition of the glass flask samplings.
 We analyzed the results of a simulation performed using a three-dimensional off-line atmospheric transport model: NIES TM [Maksyutov et al., 2008]. NIES TM was run with a resolution of 2.5° × 2.5° and 15 vertical σ levels from surface to σ = 0.03 (∼30 hPa). The advection scheme is semi-Lagrangian, and a mass fixer is adopted. Vertical mixing in the model is represented by cumulus convection and turbulent diffusion with explicitly parameterized physical processes in the planetary boundary layer. For details on NIES TM, see Maksyutov et al. .
 We used monthly CH4 flux data compiled by Patra et al.  (E2 scenario), based on the Emission Database for Global Atmospheric Research (EDGAR 32FT2000) [Olivier and Berdowski, 2001] for anthropogenic CH4, scaled to match the annual total anthropogenic emission trends of the EDGAR-HYDE (History Database of the Global Environment) database [van Aardenne et al., 2001] up to the year 2000, and based on the NASA Goddard Institute for Space Studies (GISS) emissions [Fung et al., 1991] for natural CH4 that were scaled. The aim of this analysis is to identify the effects of the IAV of atmospheric transport on the CH4 growth rate; thus, we used emissions for the year 2000, which are repeated each year. The results of the CH4 simulation are very similar regardless of whether we use repeated emissions for 2000 or trend-based emissions, because the annual total emissions varied by less than 0.3% during the analysis period (from 573.4 Tg-CH4 yr−1 in 1994 to 575.0 Tg-CH4 yr−1 in 2000) [Patra et al., 2009].
 Chemical destruction of CH4 by OH radicals was calculated up to 100 hPa by using the monthly climatology of OH radical concentrations [Spivakovsky et al., 2000] with the rate constant from the Jet Propulsion Laboratory (JPL) database [Sander et al., 2006]. Chemical destruction above 100 hPa was not considered in the NIES TM simulation; instead, the total surface emission of CH4 was scaled by a factor of 0.93 to represent the expected stratospheric loss of 40 Tg-CH4 yr−1 [Denman et al., 2007]. This factor was uniformly applied for the globe and for all months.
 The model was initialized by an average latitudinal distribution obtained from analyzed data presented by GLOBALVIEW-CH4  for January 1991. After 3 years of spin-up, the model was run from January 1994 to June 2010 with meteorological fields of the Japan Meteorological Agency (JMA) Climate Data Assimilation System (JCDAS) [Onogi et al., 2007]. In this simulation, the IAV of atmospheric transport was modeled using realistic meteorological drivers, although no IAV was given to the CH4 source or to OH concentrations. Therefore, IAV in the simulated CH4 values was caused by atmospheric transport in the model and by temperature anomalies that affect the reaction rate between OH and CH4.
3. Results and Discussion
Figure 3 shows time series of CH4 mixing ratios, their long-term trends, and growth rates derived from VOS measurements at 16 latitudes from October 1994 (from February 2003 at 30°S and 35°S) to June 2010. To show the distribution of the CH4 mixing ratios and growth rates, Figure 4 presents a set of time series of the long-term trends for all 16 latitudes and a contour plot of the growth rate in a time-latitude section.
Figure 5 shows the long-term trend and growth rate derived from simulated CH4 data. We used the daily (12Z) output of the model at the lowest model layer (from the surface to σ = 0.97, approximately 1000–970 hPa). The model data were sampled on the same date and at the nearest model grid to the longitude and latitude of the VOS measurements.
3.1. CH4 Increases During 1997–1998
 Large CH4 increases were observed in 1997–1998, in both the NH and SH. The magnitude and timing of the highest growth rate differed with latitude. In the tropics (10°N–5°S), the maximum growth rates were 19 ± 3, 18 ± 3, 16 ± 3, and 15 ± 3 ppb yr−1 at 10°N, 5°N, 0°, and 5°S, respectively, in mid-1997. In the northern midlatitudes (40°N–50°N), the maximum growth rate was 10 ± 2 ppb yr−1 at 40°N and 16 ± 1 ppb yr−1 at 50°N, in early 1998. In the southern tropics (10°S–25°S), the maximum growth rate was 22 ± 4, 16 ± 2, 15 ± 1, and 12 ± 2 ppb yr−1 at 10°S, 15°S, 20°S, and 25°S, respectively, in mid-1998. At 20°N and 25°N, the growth rate was relatively low during 1997–1998, being 6 ± 5 ppb yr−1. The timing of the onset and maximum of growth in each region did not change when we applied a digital filter with a different cutoff period (e.g., 21 months).
 Previous global analyses using global measurement networks showed that the maximum CH4 increase (15 ppb yr−1) was observed in 1998 [Dlugokencky et al., 2001, 2009; Langenfelds et al., 2002; Simpson et al., 2002; Rigby et al., 2008]. These results are in good quantitative agreement with our CH4 growth rates for 1998, as observed in the northern midlatitudes and in the SH. We found an interhemispheric difference in the timing of the 1998 growth: early 1998 in the northern midlatitudes and mid-1998 in the SH, which is consistent with a previous study [Langenfelds et al., 2002]. However, the CH4 growth rates in the northern tropics during 1997, based on our measurements in the western Pacific, are much larger than those derived from other global CH4 data sets (about 5 ppb yr−1 in mid-1997 and 10 ppb yr−1 in late 1997) [Dlugokencky et al., 2001, 2009; Langenfelds et al., 2002; Simpson et al., 2002; Rigby et al., 2008]. These indicate that the observed signal in the northern tropics can be attributed to regional variations, although large parts of the observed signal in the northern midlatitudes and in the SH are consistent with the global trend (see section 3.3).
 The present simulation did not reproduce the CH4 increases observed in 1998, indicating that an explanation of the increases requires an interannual decrease in CH4 sinks [e.g., Manning et al., 2005] and/or an increase in surface emissions, such as enhanced emissions from biomass burning in 1997–1998 [van der Werf et al., 2006] and from wetlands due to anomalously high temperatures in 1998 [Dlugokencky et al., 2001; Langenfelds et al., 2002]. The simulation was successful in reproducing the large CH4 increase over the northern tropics in mid-1997.
 From 1999 to 2006, the CH4 increase was approximately zero at all latitudes. We observed a CH4 decrease during 1999–2000 and 2003–2005 at all latitudes. A moderate CH4 increase of 5 to 9 (±2–4) ppb yr−1 was observed in the NH during the El Niño event of 2002–2003. A discrepancy between simulation and observation data is seen from 1999 to 2006, as the lack of a leveling-off in the trends of simulated CH4. This discrepancy may be attributed to the IAV and trends in OH concentrations and surface CH4 emissions that were excluded from the simulation.
3.2. CH4 Increases in 2007
 We observed CH4 increases again in 2007, although not between the Equator and 20°N. The growth rate in 2007 was higher in the NH than in the SH. The maximum growth rate was observed in mid-2007: 10 ± 1, 12 ± 1, 9 ± 1, 10 ± 2, and 7–8 ± 1 ppb yr−1, at 50°N, 40°N, 5°S, 10°S, and 15°S–35°S, respectively. The results show that CH4 increase has occurred since mid-2006 at 25°N and 30°N (the maximum growth rate was 12 ± 3 ppb yr−1 in late 2006). These growth rates for 2007 are the second largest after those for 1997–1998. However, a CH4 increase was not found for the northern tropics; the growth rates in mid-2007 were −2 ± 1 (10°N), 0 ± 2 (15°N), 1 ± 1 (5°N), and 5 ± 1–2 (0° and 20°N) ppb yr−1. In the northern tropics, the seasonal minimum values of CH4 mixing ratio in the summer showed an increase during 2007–2008, whereas the maximum values in the winter decreased, resulting in the lack of an increasing trend and a reduced amplitude of the seasonal cycle.
 The high rates of CH4 growth diminished after 2007. The growth rates were 1 ± 1 (40°N–50°N), 3–5 ± 3 (20°N–30°N) and 3–5 ± 1–2 (SH) ppb yr−1 in 2009. In contrast, large CH4 growth in 2009 was observed only over the northern tropics, which yielded values of 13 ± 2, 17 ± 1, 12 ± 2, and 10 ± 1 ppb yr−1 at 15°N, 10°N, 5°N, and 0°, respectively.
 The growth rate in 2007 differed between the hemispheres. Assuming no change in the OH concentration, an increase in global surface emissions by ∼20 ± 3 Tg-CH4 yr−1 is required to explain the observed CH4 growth rate of 7 ± 1 ppb yr−1 in the SH. A larger increase in surface emissions in the NH (by 4–7 ± 3 Tg-CH4 yr−1) compared with the SH is required to explain the higher growth rate (by 3–5 ± 1 ppb yr−1) in the NH than in the SH. Regarding the effect of OH variations on the CH4 increases in 2007, our simulation showed that a global OH decrease of 4.5 ± 0.6% yr−1 caused a CH4 increase of 7 ± 1 ppb yr−1 in the SH and 10 ± 1 ppb yr−1 in the NH, in the case of no changes in surface emissions. This variation in OH concentrations was able to explain the NH–SH difference in the CH4 growth rate (3 ppb yr−1), because of the NH–SH difference in the CH4 and OH mixing ratios, and meteorological parameters. To explain the higher CH4 growth rate of 12 ± 1 ppb yr−1 in the NH, an OH decrease of 5.5 ± 0.5% yr−1 is required. The simulated OH decreases of 4.5–5.5% yr−1 in 2007 are consistent with an OH decrease of 4 ± 14% from 2006 to 2007, as estimated from measurements of methyl chloroform [Rigby et al., 2008].
 The simulation reproduced regional differences in the CH4 growth rate during 2007, i.e., a significant decrease in the northern tropics. This finding indicates that the suppression of CH4 increase in 2007 within the northern tropics, as well as the large CH4 increase in mid-1997, may be attributed to the effects of IAV in atmospheric transport.
 To illustrate the spatial differences in CH4 variations in 1997 and 2007, we show the spatial distributions of the growth rate of simulated CH4 at the surface in April 1997 (Figure 6a) and in January 2007 (Figure 6b), when the maximum and the minimum growth rates were found in the northern tropics of the western Pacific, respectively. We used the monthly mean output of the model for the global analysis shown in Figure 6, and the growth rate was calculated at each model grid point by the same method. Also shown are monthly anomalies (i.e., deseasonalized) of horizontal surface winds. Monthly anomalies are the difference between a given monthly mean and the average of all monthly means for that calendar month over the period 1994–2009. The simulated CH4 growth rates show a spatially inhomogeneous pattern in the NH. The results indicate that the CH4 increase in 1997 and the CH4 decrease in 2007 were regional events that occurred in the northern tropics over the western Pacific. In other regions, small areas of CH4 increases and decreases were found in 1997 and 2007.
 The surface wind field over the western tropical Pacific shows large differences between April 1997 and January 2007. The high growth rates in 1997 (10 ppb yr−1 or more) were accompanied by strong westerly wind anomalies that are a well-known feature related to El Niño events. In 1997, winds in the equatorial Pacific were actually from the west (rather than typical easterlies), not only westerly anomalies. This finding suggests that the CH4 increase in 1997 over the western tropical Pacific resulted from the advection of CH4-rich air from Southeast Asia, where large CH4 sources are found. The observed and simulated CO2 data show no clear increase in CO2 over 0°–15°N in mid-1997 because CO2 sinks occur in tropical ecosystems, but not sources.
 In 1997, massive forest fires in Indonesia influenced tropospheric chemistry [e.g., Duncan et al., 2003]. Large decreases in OH concentrations resulting from the Indonesian fires were found from measurements of methyl chloroform [Prinn et al., 2005] and carbon monoxide containing radiocarbon [Manning et al., 2005]. The global enhancement of CH4 mixing ratio observed in late 1997 was probably linked to the reduced OH, and CH4 directly released from the biomass burning emissions, as reported in previous studies [Dlugokencky et al., 2001, 2009; Langenfelds et al., 2002; Simpson et al., 2006; Rigby et al., 2008]. Our observational data (not simulation results) for the western Pacific would certainty have captured the CH4 anomaly in 1997 associated with the Indonesian fires, at the correct latitudes and timing and with a higher magnitude than those derived from global data sets. This CH4 anomaly would propagate globally and cause CH4 increases during late 1997 and 1998.
 The suppression of CH4 growth in 2007 over the northern tropics was also influenced by atmospheric transport, possibly associated with vertical transport. In the western tropical Pacific, especially during La Niña events, deep cumulus convections associated with a strong Walker circulation are observed in this region [e.g., Diamond and Bell, 2008]. The deep convection lifts the CH4-rich air from the surface to the upper troposphere [Patra et al., 2009]. This idea is supported by our simulation result, which shows a CH4 increase at 200 hPa, slightly east of the region of CH4 decrease at the surface. The surface wind anomalies indicate that CH4-poor air was transported from the eastern Pacific to the western tropical Pacific by easterly wind anomalies. These findings indicate that the IAV in the meteorological field, mainly atmospheric circulation associated with La Niña, caused the suppression of CH4 growth in 2007 within the western tropical Pacific. A similar suppression of the CH4 growth rate was observed and simulated in this region during La Niña events in 1995–1996 and 1999–2000 (Figures 4b and 5b).
 Temperature anomalies have an influence on the reaction rate of OH with CH4. Higher-temperature anomalies result in higher reaction rates and hence a larger CH4 sink. We did not consider variations in OH concentrations. Large decreases in CH4 sinks (up to −5 ppb) were simulated in April 1997, corresponding to strongly negative temperature anomalies (up to −5 K); e.g., Europe, the Middle East, and South Asia (Figures 6c and 6e). In January 2007, temperature anomalies had only a minor influence on CH4 sinks throughout the NH, because OH concentrations were low during winter (Figures 6d and 6f).
 In the western tropical Pacific, temperature anomalies were very small: −0.02 K in April 1997 and 0.15 K in January 2007, as obtained by averaging the values for 120–180°E, 0–20°N. Consequently, the monthly anomalies of CH4 sinks were also very small in the western tropical Pacific (−0.08 ppb in April 1997 and 0.03 ppb in January 2007), as were the annually accumulated anomalies of CH4 sinks (−0.3 ppb yr−1 for October 1996 to October 1997; 0.6 ppb yr−1 for July 2006 to July 2007). The sign of the CH4 sink anomalies (negative in 1997 and positive in 2007) is consistent with that of CH4 growth rates; however, the value of CH4 sink anomalies is too small to explain the simulated growth rate of 10–20 ppb yr−1 in 1997, and explains about 20% of the simulated growth rate of −3 ppb yr−1 in 2007.
 We presented an analysis of IAV and trends in atmospheric CH4 mixing ratio from 55°N to 35°S over the western Pacific from 1994 to 2010, observed quasi-monthly (∼10 times per year) with high latitudinal resolution by NIES/CGER using VOS. Large CH4 increases were observed during the El Niño event of 1997–1998: the maximum growth rates in 1998 were 10–16 (±1–2) ppb yr−1 in northern midlatitudes and 12–22 (±1–4) ppb yr−1 in the SH. The CH4 increase leveled off at all latitudes from 1999 to 2006, although a moderate CH4 increase of 5–9 (±2–4) ppb yr−1 was observed in the NH during 2002–2003. The CH4 mixing ratios increased again in 2007, with maximum values of 10–12 (±1–3) ppb yr−1 at 25°N–50°N and 7–8 (±1) ppb yr−1 at 15°S–35°S. These trends are quantitatively consistent with the results obtained from global measurement networks [Dlugokencky et al., 2001, 2009; Langenfelds et al., 2002; Simpson et al., 2006; Rigby et al., 2008], except for the tropics. The high growth rates of CH4 diminished after 2007, although large CH4 growth (10–17 (±1–2) ppb yr−1) was observed within the northern tropics (0°–15°N) in 2009.
 Regarding the CH4 increase in 2007, our observations indicate that an increase in global surface emissions of ∼20 ± 3 Tg-CH4 yr−1, combined with a larger increase (by 4–7 ± 3 Tg-CH4 yr−1) in the NH than in the SH, is required to explain the increase, assuming no change in OH concentrations. The results of simulations using the NIES TM global atmospheric transport model indicate that a global OH decrease of 4.5 ± 0.6%–5.5 ± 0.5% yr−1 is required in the case of no changes in surface emissions of CH4.
 We found two remarkable phenomena in the IAV of CH4 in the northern tropics (around 0° to 15°N) over the western Pacific: a high growth rate of 15–19 ± 3 ppb yr−1 in mid-1997, ahead of the global increase in 1998, and the suppression of CH4 growth in 2007. We used NIES TM to investigate the effects of IAV in the meteorological field on the CH4 growth rate in 1997 and 2007. The simulation we performed here did not consider the IAV in CH4 sources and sinks; however, it reproduced the CH4 increase in 1997 and the suppression of CH4 growth in 2007 over the northern tropics. We conclude that the enhanced CH4 mixing ratios observed in 1997 were transported from Southeast Asia due to El Niño–related wind anomalies. The suppression of the CH4 growth rate in 2007, as well as in other La Niña years (1995–1996 and 1999–2000), may have been associated with the anomalously strong vertical transport.
 We observed large El Niño/La Niña–related variability in atmospheric circulation within the northern tropics of the western Pacific. Furthermore, our observations were made at sites located relatively close to the large CH4 sources of East and Southeast Asia, which resulted in the high sensitivity of measured CH4 mixing ratios in the northern tropics to changes in atmospheric transport and emissions from East and Southeast Asia. The IAV of atmospheric CH4 as indicated by our observations in the western Pacific represent a superposition of the global features of background CH4 in the northern midlatitudes and the SH, and in particular the local features of CH4 variations in the northern tropics. The CH4 data set presented here would be valuable in accurately and quantitatively estimating the global CH4 budget (especially, emissions related to biomass burning in Indonesia during 1997) in a future study.
 We are grateful to Shigeru Kariya, Tomoyasu Yamada, Tomoko Nojiri, the staff of the Global Environmental Forum, and the companies and crew of the M/S Hakuba Maru, the Southern Cross Maru, the Golden Wattle, the MOL Glory, the FujiTrans World, the Transfuture 5, the Skaugran, the Alligator Hope, the Pyxis, and the Skaubryn for their continuous support in collecting air samples. We would also like to thank Prabir K. Patra for providing CH4 flux data, Keiichi Katsumata for his work on the NIES CH4 scale, Hisayo Sandanbata and Yoko Kajita for their assistance with CH4 measurements and flask preparation, and Takakiyo Nakazawa and Misa Ishizawa for providing the source code for data fitting. The JCDAS data sets used for this study are provided by the cooperative research project of the JRA-25 long-term reanalysis by the Japan Meteorological Agency (JMA) and the Central Research Institute of Electric Power Industry (CRIEPI). The NIES TM simulation was performed using the NIES supercomputer system (NEC SX-8R/128M16).