GOSAT is a joint project of the Japanese Ministry of the Environment (MOE), the National Institute for Environmental Studies (NIES) and the Japan Aerospace Exploration Agency (JAXA) to monitor the global distribution of atmospheric CO2 and CH4 from space (Yokota et al., 2009). The retrieved XCH4, as a part of NIES GOSAT Level 2 (L2) product (v02.xx), has been reported to have a mean bias of -5.9 ppb and mean standard deviation of 12.6 ppb against the XCH4 at selected TCCON sites (Yoshida et al., 2013). In this study, we analysed NIES GOSAT L2 XCH4 (v02.21) without any bias correction. The latest data processing and the auxiliary information are described in GOSAT User Interface Gateway, http://data.gosat.nies.go.jp/GosatUserInterfaceGateway/guig/doc/documents/doc_en_docdist.html.

We analysed GOSAT XCH4 over two regions in northeast Asia separately (Fig. 1). One is over northeastern China–Korea (115–130∘ E, 30–40∘ N), and the other is over Japan (130–145∘ E, 30–40∘ N). The northeastern China–Korea region covers highly populated and industrialized areas with large anthropogenic CH4 sources in the Eurasia continent. The Japan region has small CH4 sources, but located downwind of the continental CH4 emissions. Time series of XCH4 data from June 2009 to March 2014 over the two regions with monthly means are shown in Fig. 2. It is noted that we used only the XCH4 over land to minimize possible errors depending on sounding observation mode (Fig. 3). In fact, since a few soundings over ocean around East Asia were retrieved, removed XCH4 data through this criterion are less than 5 % of the total. A long-term trend component in each XCH4 dataset derived through a digital filtering of two-year cutoff period (Nakazawa et al., 1997) is also plotted in Fig. 2. To focus the seasonal variations, the trend components were removed, and the detrended XCH4 time series are further analysed.

The GOSAT XCH4 retrievals over northeastern China–Korea have clear seasonality with high peaks in summer and low peaks in winter. The summertime high XCH4 appear to be influenced by the seasonal biogenic CH4 emissions from rice paddies and natural wetlands underneath in east China and Korea. The summer peak in 2013 was more prominent than the preceding two years, 2011 and 2012. Also, the summertime XCH4 retrievals over northeastern China–Korea in 2009 and 2010 were relatively high while no significantly high XCH4 was observed over Japan. Since there is a limited number of retrieval available over Japan for the first two years of the GOSAT operation, it is difficult to discuss the XCH4 difference over the two regions for 2009 and 2010. We thus leave this topic for a future investigation.

The seasonality of the GOSAT XCH4 retrieval over Japan is overall similar to northeastern China–Korea. Although the seasonal cycle varies largely year-to-year, XCH4 retrievals of August and September in 2013 were outstandingly high. Japan is located downwind of strong anthropogenic and natural biogenic CH4 emissions in the continent, and then the signals of the continental CH4 emissions are decreased as the air is transported. However, it is noticeable that, in the summer of 2013, the XCH4 retrievals over both Japan and northeastern China–Korea regions reached the almost same high levels. This comparable XCH4 levels in the two regions indicates there was a mechanism of fast atmospheric transport in 2013 to bring CH4-rich air to Japan with less diffusion than the preceding years.

The number of GOSAT retrievals over Japan increased in 2013 and 2014, compared with those in the previous years. This increase is due to the change of the observation strategy to increase available GOSAT retrievals over Japan. The initial regular schedule, there were fewer soundings over lands, but most soundings were over oceans or land-ocean mixed locations. The soundings over ocean or mixed locations are difficult to be retrieved. As a result, a few retrievals remained over Japan after screening. Aiming at increasing the retrievals over Japan, the observation locations were moved inland from ocean and mixed locations as much as possible. The observation strategy change was made as a concerted decision by GOSAT Project terms among the three agencies NIES, JAXA, and MOE. This observation change was implemented on 6 May 2013.

The spatial maps of GOSAT retrievals we used in this study are shown in Fig. 3. As seen in Fig. 3a, most of the soundings were taken at lower surface elevation; more than 80 % is below 100 m, 95 % is below 500 m. The spatial maps of the retrievals in August and September in 2012 and 2013 are shown in Fig. 3b and c, respectively. Kort et al. (2014) applied the elevation correction to SCIAMACHY XCH4 over in the southwestern USA. The elevation impact is mainly exhibited over the Rocky Mountains because of low CH4 air in the stratosphere. Surface elevation/topography could be one of potential biases in GOSAT XCH4. However, for the GOSAT data we used, we found no statistically significant correlation with surface elevation (< r = -0.11) and also the GOSAT XCH4 data sampled at elevation of more than 1000 m is a few (∼ 0.1 % out of total). Therefore, no elevation correction was applied to the GOSAT XCH4 in this study. In Fig. 3b and c, the observation strategy change mentioned above is noticeable, the coverage of retrievals over Japan was dramatically increased in 2013, compared with 2012. Regarding China and Korea, there is no significant difference between 2012 and 2013.

Inside the Japan region of this study, ground-based XCH4 measurements have been conducted at two TCCON sites, Saga (33.24∘ N, 130.29∘ E) and Tsukuba (36.05∘ N, 140.12∘ E) as shown in Fig. 1. TCCON is a worldwide network of ground-based high-resolution FTSs, which record spectra of the direct sunlight in the near-infrared, and provides accurate and precise column-averaged dry-mole fractions of atmospheric constituents including CO2, CH4, N2O, HF, CO, H2O, and HDO retrieved from these spectra absorbed by them (Wunch et al., 2011). The TCCON XCH4 measurements have an estimated uncertainty of 7 ppb (2σ) (Wunch et al., 2010). TCCON data play a critical role in the validation of space-based measurements. The Saga TCCON site is in Kyushu Island, operated by JAXA since June 2011. The Tsukuba TCCON site is located ∼ 50 km north of Tokyo in the Japan main island, operated by NIES since 2009. These two Japanese TCCON sites are apart ∼ 1000 km longitudinally. In this study, we use the TCCON data processed by GGG 2012.

Figure 4 shows XCH4 retrievals at Saga and Tsukuba TCCON sites during the period for 2011 to 2014. We processed the both TCCON XCH4 time series in the same manner with the GOSAT XCH4 to obtain the long-term trends that are shown in blue lines in Fig. 4. It is interesting that, before the summer 2013, XCH4 retrievals at Tsukuba overall are lower than at Saga. Since Saga is located closer to the continent than Tsukuba, Saga is considered to be influenced by the continental anthropogenic CH4 emissions more strongly than Tsukuba. In the summer of 2013, extremely high XCH4 retrievals both at Saga and Tsukuba were observed, reaching almost a same level. This XCH4 enhancement observed at the ground-based TCCON sites is coincident with the high XCH4 observed by GOSAT, and strongly supports our speculation that the CH4 rich air was transported quickly from the continent to Japan for this period.

To focus on the seasonal and synoptic variations, we compared the detrended XCH4 time series from GOSAT over Japan and the two Japanese TCCON sites. Figure 5a shows that all the detrended XCH4 data are overall in phase of seasonal cycle with seasonal amplitude of ∼ 20 ppb. Compared with TCCON XCH4, GOSAT XCH4 shows large short-term variability. In 2013, both GOSAT and TCCON XCH4 together rapidly increased in August and remained high in September. In 2012, both GOSAT XCH4 and TCCON XCH4 show no clear tendency in August, and all of them appear to be upward in September. On average, the XCH4 level of GOSAT over Japan in August and September 2013 is higher by ∼ 15 ppb than 2012. The XCH4 levels of both TCCON sites in 2013 are higher by ∼ 20 ppb than 2012. These enhancements of XCH4 are comparable to their seasonal amplitude.

To examine further how the synoptic variability of GOSAT is correlated with TCCON, we removed the mean seasonal cycles from the detrended XCH4 time series and took the monthly means (Fig. 5b). Except the months when the retrievals are available for less than 2 days, the correlation coefficients (r) of the monthly means between GOSAT and TCCON at Saga, and between GOSAT and TCCON at Tsukuba, are 0.81 and 0.61, respectively. These correlation coefficient values exceed the 95 % significance level. Despite the large short-term variability, the synoptic variability of GOSAT over Japan is overall correlated with the TCCON XCH4 at two Japanese sites. The enhancement of XCH4 in the summer of 2013 is consistent among GOSAT and TCCON. If the period is limited to May–December 2013, when the number of GOSAT XCH4 retrievals was increased due to the observation strategy change mentioned earlier, the correlation coefficients (r) between GOSAT and TCCON are improved to be 0.91 with Saga and 0.96 with Tsukuba. This implies that the increase in the observations over Japan improves the capability of GOSAT to detect synoptic variability in XCH4.

In order to see the relationship between the surface CH4 concentration and the enhancement of GOSAT XCH4 over Japan, we analyzed the surface CH4 concentrations observed at three ground-based monitoring stations in Japan, Cape Ochi-ishi (COI, 43.16∘ N, 145.49∘ E), Ryori (RYO, 39.03∘ N, 141.82∘ E), and Yonagunijima (YON, 24.47∘ N, 123.02∘ E). These site locations are shown in Fig. 1. At all the stations, continuous measurements of atmospheric CH4 are conducted. Cape Ochi-ishi (COI) is a station operated by NIES, which is located at the east tip of Hokkaido island (Tohjima et al., 2002). Ryori (RYO) is located inside the Japan region defined in this study, where the monitoring of surface greenhouse gas concentrations has been conducted by the Japan Meteorological Agency (JMA) as a part of the Global Atmospheric Watch (GAW) program of the World Meteorological Observation (WMO). RYO is on the west coast of the Japan main island, about 300 km north of Tsukuba and far away from direct influences of residential and industrial pollutants. Yonagunijima (YON) is also one of JMA-operated GAW stations, which is located far south of the Japan main island and east of ∼ 110 km of Taiwan. The detailed station descriptions on RYO and YON are found on the JMA website (http://www.data.jma.go.jp/gmd/env/ghg_obs/en/station).

The time series of surface CH4 concentrations at the three ground-based stations are shown in Fig. 6, with their monthly means and long-term trends. Here we analyzed the afternoon mean CH4 (averaged hourly CH4 over 12:00–15:00 local time) from the respective data sets, assuming that the afternoon values are large-scale representative. The observed CH4 concentrations at all the sites show similar seasonal cycles in timing. Seasonally the CH4 values are low in July and August, and high in winter to spring. In the winter, the westerly wind prevails and transports the CH4-rich air from the continent (mainly anthropogenic CH4 emitted in east China) to Japan, causing the rise of CH4 concentrations. In the summer, the southeasterly wind is dominant, bringing clean air to Japan from the Pacific Ocean, where there is no major CH4 source, so that the surface CH4 concentrations become low.

In the summer of 2013, unseasonably high CH4 concentrations were observed at RYO with a sharp increase in the middle of August. The CH4 concentrations at COI started increasing earlier from its summer minimum than the previous year, 2012. At YON, no significant CH4 enhancement was seen in 2013 compared with the previous years. Since no similar CH4 change to RYO and COI was observed at YON, the farthest southwestern island of Japan, this significant CH4 enhancement event appears to be spatially limited in the area around Japan main island and Hokkaido island. To further examine the summer increase of surface CH4 concentrations, we compared the detrended CH4 at RYO and COI for the two years of 2012 and 2013 (Fig. 7). The timing and amplitude of seasonal cycles at RYO and COI overall agree well with each other, except for the summer of 2013. In August and September of 2013, the temporal variations of CH4 at RYO and COI are different from those in the previous year 2012, when the CH4 concentrations were low over the summertime and started rising at the end of September. In August 2013, the abrupt CH4 increase by ∼ 100 ppb was observed at RYO, followed by COI with ∼ 1 week delay. In September, the CH4 at both sites decreased but stayed in the higher level than 2012. Given that the fact the major CH4 sources in east China, the sudden large increase of CH4 in August 2013 is probably caused by unseasonal transport of CH4-rich air from the continent to Japan though normally in August the wind with CH4-low air from the Pacific Ocean is prevailing over Japan.

Forward modelling gives us insights into the contribution of atmospheric transport on the enhancement of XCH4 and surface CH4 concentration in the summer of 2013 over Japan. Here we examine the 2013 summertime atmospheric transport over the northeastern Asia.

Japan's summer climate is governed by the Pacific High (a lower-level high-pressure system) and the Tibetan High (an upper-level high-pressure system). These pressure systems were reported to have been enhanced during July and August 2013 (Tokyo Climate Center News No.34 Autumn 2013, available at http://ds.data.jma.go.jp/tcc/tcc/news). The Pacific High continued to expand westward and largely developed over the western part of Japanese islands including Okinawa. The Tibetan High expanded to the Japan main island in line with the northward meandering of upper-level westerly winds (the subtropical jet stream). The enhanced atmospheric transport from east China to Japan was probably attributed to those anomalously developed high-pressure systems.

To see how the 2013 summertime atmospheric transport differs from the mean transport pattern, Fig. 11 shows the wind fields at the surface and at 850 hPa pressure level, from the JCDAS wind fields of August and September in 2013 over East Asia, compared with those of the mean wind fields for the five years of 2009–2013.

At the surface level (Fig. 11a), the mean wind field clearly shows that, in August the southeasterly wind from the Pacific Ocean prevails due to the development of the Pacific High. In September the wind from the continent to Japan start blowing as the Pacific High is retiring. In August 2013, as the Pacific High expanded westward, the air moved northward along the coast of China, turned around the Korean Peninsula, and flowed to Japan. This wind pattern suggests that the CH4-rich air was transported from east China to Japan in 2013, while the clean air is normally transported from the Pacific Ocean. In September 2013, over the Pacific Ocean, south of the Japan main island, easterly wind was still stronger than the normal, but the wind pattern over Japan was almost back to the normal, which can be characterized as a weak convergence of westerly wind from the continent and easterly wind from the Pacific Ocean. This nearly normal wind pattern over northern Japan would lower the CH4 concentrations at the surface level as observed at RYO and COI.

At the 850 hPa level (Fig. 11b), it is notable that, in August 2013 the air moved over east China along the coast and turned around the Korean Peninsula sharply to the Japan archipelago. The anomalous westerly winds were stronger in the upper levels than near the surface. Given the major CH4 source distributions in East Asia, the strong northward air flow along the coast could reduce local CH4 accumulation, but transport the CH4-rich air effectively to the north and then to Japan as turning around the Korean peninsula. In September 2013, the wind speed over Japan was much lower than August, but wind still blows westerly from the continent to Japan. This slower westerly air flow could maintain the higher level of XCH4 over Japan during the September 2013.

The wind patterns we examined above shows us how the atmospheric transport field in 2013 differed from a climatological field on a regional-scale. In order to narrow down the origins and the upstream patterns of the air masses to the Japanese TCCON sites, we conducted back trajectory analysis using the Stochastic Time-Inverted Lagrangian Transport (STILT) model (Lin et al., 2003), driven by Global Data Assimilation System (GDAS) meteorology (1∘ × 1∘). To obtain the monthly mean features of the upstream, we released 100 particles from the height of 1500 m (approximately ∼ 850 hPa) at Saga and Tsukuba, at every 12:00 noon local time (= 03:00 UT) and travelled backward for 10 days. Every 30 min, the number of particles was counted by a 1∘ × 1∘ air column and the total number of particles over the 10 day duration was divided by the maximum number per column. Thus, we obtained a normalized daily upstream pattern and averaged them over a month. Figure 12 shows the monthly normalized trajectories for August and September in 2012 and 2013. There are distinct differences in the upstream patterns between 2012 and 2013. The patterns of the summer of 2012 are almost climatological; in August, the wind flows dominantly from the Pacific to the Japan, in September the dominant wind direction is in transition; from southeasterly wind (from the Pacific) to northwesterly wind (from the continent). On the other hand, in August 2013 the air masses reached the Japanese TCCON sites from the west, after travelling over the coastal side of east China. In September 2013, the westerly wind from the continent is still dominant, especially for Saga. This backtrajectory result supports that the anomalous wind field in the summer of 2013 brought the CH4-rich air from China to Japan, resulting in the high XCH4 observed at the two Japanese TCCON sites and also by GOSAT over Japan.

Although we suggest that the atmospheric transport field probably attributes to the enhancement of XCH4 and CH4 concentration observed in the summer of 2013, we cannot entirely rule out other possible factors. Here we discuss two factors. One is the surface emission changes. Though the temporal variations in XCH4 do not necessarily correlate with the surface emissions (e.g. Bloom et al., 2010), the surface emission change is potential to impact on the change in XCH4. The second is the contribution of stratospheric methane.

The CH4 emissions from rice cultivations and wetland in southern China might be enhanced under the hot summer condition in 2013. East Asia around China experienced a hotter summer monsoon season (June–September) by more than 1 ∘C than the season normal (Tokyo Climate Center News No.34 Autumn 2013), while less than 60 % of the normal precipitation in eastern China was reported. A hot weather condition increases the CH4 emissions through the enhancement of photosynthesis and methanogenic activity in inundated grounds such as wetlands and rice paddies; while a dry condition reduces the CH4 emissions from wetlands as the water table levels in the ground become low. Thus, the hot and dry weather conditions have opposite effects on the CH4 emissions from wetlands. The time delay in the correlation between CH4 emissions and climate anomalies should be considered as the groundwater plays an important role in wetland CH4 emissions. Furthermore, since rice cultivation is human-managed, multiple controlling factors on CH4 emissions from rice paddies should be considered. A further investigation of wetland and rice CH4 emission changes responding to the climate anomaly in East Asia is needed.

Another possibility is the contribution of stratospheric methane. Saad et al. (2014) presented the analysis that the stratospheric methane causes short-term fractionations in total column averaged CH4 observed at several TCCON sites. The contribution of stratospheric methane to the anomaly in summer 2013 is supposed to be minor or less influential. Firstly the surface CH4 concentrations at COI and RYO increased in August 2013 when the XCH4 anomaly occurred, suggesting the major contributor on the anomaly is in the troposphere. Secondly, the order of the stratospheric methane fractionation is smaller than ∼ 3 ppb, which would not be enough to produce the anomaly of an order of ∼ 20 ppb.