The Greenhouse gases Observing SATellite (GOSAT) (also referred to as Ibuki) project is developed jointly by the National Institute for Environmental Studies (NIES), Ministry of the Environment (MOE), and Japan Aerospace Exploration Agency (JAXA). It has been providing columnar dry-air mole fractions of the two important greenhouse gases (XCH4 and XCO2) at near-global coverage since its launch in January 2009. It is equipped onboard with the Thermal And Near infrared Sensor for carbon Observation-Fourier Transform Spectrometer (TANSO-FTS) and the Cloud and Aerosol Imager (TANSO-CAI) (Kuze et al., 2009). To avoid cloud contamination in the retrieval process, any scene with more than 1 cloudy pixel within the TANSO-FTS IFOV is excluded. The atmospheric images from CAI are used to identify the cloudy pixels. As a result of this strict screening, only limited numbers of XCH4 data are available during the SW monsoon over South Asia. This study uses the GOSAT SWIR XCH4 (Version 2.21)-Research Announcement product for the period of 2011–2014. The ground-based FTS measurements of XCH4 by the Total Carbon Column Observing Network (TCCON) (Wunch et al., 2011) are used extensively to validate the GOSAT retrievals. Retrieval bias and precision of column abundance from GOSAT SWIR observations have been estimated as approximately 15–20 ppb and 1 %, respectively, for the NIES product using TCCON data (Morino et al., 2011; Yoshida et al., 2013).

Model analysis is comprised of simulations from JAMSTEC's atmospheric general circulation model (AGCM)-based chemistry-transport model (ACTM; Patra et al., 2009). The AGCM was developed by the Center for Climate System Research/National Institute for Environmental Studies/Frontier Research Center for Global Change (CCSR/NIES/FRCGC). It has been part of transport model intercomparison experiment TransCom-CH4 (Patra et al., 2011a) and used in inverse modeling of CH4 emissions from in situ observations (Patra et al., 2016). The ACTM runs at a horizontal resolution of T42 spectral truncations (∼ 2.8∘×2.8∘) with 67 sigma-pressure vertical levels. The evolution of CH4 at different longitude (x), latitude (y), and altitude (z) with time in the Earth's atmosphere depends on the surface emission, chemical loss, and transport, which can be mathematically represented by the following continuity equation: dCH4x,y,z,tdt=SCH4x,y,t-LCH4x,y,z,t-∇ϕx,y,z,t, where CH4 is the methane burden in the atmosphere, SCH4 is the total emissions/sinks of CH4 at the surface, LCH4 is the total loss of CH4 in the atmosphere due to the chemical reactions, and ∇ϕ is the transport of CH4 due to the advection, convection, and diffusion.

The meteorological fields of ACTM are nudged with reanalysis data from the Japan Meteorological Agency, version JRA-25 (Onogi et al., 2007). The model uses an optimal OH field (Patra et al., 2014) based on a scaled version of the seasonally varying OH field (Spivakovsky et al., 2000). The a priori anthropogenic emissions are from the Emission Database for Global Atmospheric Research (EDGAR) v4.2 FT2010 (http://edgar.jrc.ec.europa.eu). The model sensitivity for emission is examined by two cases of emission scenarios based on different combinations of sectoral emissions. The first one is referred to as the “AGS”, where all emission sectors in EDGAR42FT are kept at a constant value for 2000, except for emissions from agriculture soils. The second one is a controlled emission scenario referred to as “CTL”, which is based on the ensemble of the anthropogenic emissions from EDGAR32FT (as in Patra et al., 2011a), wetland and biomass burning emissions from Fung et al. (1991), and rice paddy emission from Yan et al. (2009). The emission seasonality differs substantially between the CTL case and the AGS case due to differences in emissions from wetlands, rice paddies, and biomass burning; other anthropogenic emissions do not contain seasonal variations (Patra et al., 2016). Further details about the model and these emission scenarios can be found in the previous studies (Patra et al., 2009, 2011a, 2016).

XCH4 is calculated from the ACTM profile using the following equations: XCH4=∑n=160CH4(n)×Δσp(n), where CH4(n) is the dry-air mole fraction at model mid-point level, n= number of vertical sigma pressure layers of ACTM (=1–60 with σp values of 1.0 and 0.005), and Δσp= thickness of the sigma pressure level. Note here that we have not incorporated convolution of model profiles with retrieval a priori and averaging kernels. Because the averaging kernels are nearly constant in the troposphere (Yoshida et al., 2011), this approximation does not lead to serious errors in constructing the model XCH4. For both the CTL and AGS cases, we adjust a constant offset of 20 ppb to the modeled time series, which should make the a priori correction have a lesser impact on the model XCH4. Because the focus of this study is seasonal and spatial variations in XCH4, a constant offset adjustment should not affect the main conclusions.

This section presents an analysis of XCH4 observed by GOSAT from January 2011 to December 2014 over the Indian region. We characterize the four seasons specific to the region as winter (January–March), spring (April–June), summer (July–September) or the SW monsoon, and autumn (October–December), as commonly used in meteorological studies (e.g., Rao, 1976). To study the seasonal XCH4 pattern in detail depending on the distinct spatial pattern of surface emissions and XCH4 mixing ratios shown in Fig. 1, the Indian landmass was partitioned into eight sub-regions: Northeast India (NEI), Eastern India (EI), Eastern IGP (EIGP), Western IGP (WIGP), Central India (CI), Arid India (AI), Western India (WI), Southern Peninsula (SP), and two surrounding oceanic regions, the Arabian Sea (AS) and the Bay of Bengal (BOB) (Fig. 2a). Regional divisions are made based on spatial patterns of emission and XCH4 (Fig. 1a1–c2) and our knowledge of seasonal meteorological conditions. Since general features of XCH4 simulated by ACTM using emission scenarios AGS and CTL are similar to each other, the main discussion is carried out using the AGS scenario only.

Figure 1a1–a2 show that the XCH4 mixing ratios are lower in spring and higher in autumn. A strong latitudinal gradient in XCH4 is observed between the Indo-Gangetic Plain (IGP) and the other parts of India. XCH4 shows the highest value (∼ 1880 ppb) over the IGP, eastern, and northeastern Indian regions. As seen from Fig. 1b1–b2, ACTM simulations are able to reproduce the observed latitudinal and seasonal gradients in XCH4, i.e., higher values during the southwestern monsoon and autumn seasons and lower values during the winter and spring seasons over the IGP region. The optimized total CH4 fluxes (AGS and CTL) show high emissions over the IGP and northeastern Indian regions (Fig. 1c1–c2). Most elevated levels of XCH4 are often observed simultaneously with the higher emissions, suggesting a link between the enhanced XCH4 and high surface emissions in summer. However, this link is not valid for all locations. For example, over the western and southern regions of India, XCH4 is higher in autumn than in spring, though the emissions are higher in spring.

Figure 2b–k show ACTM–GOSAT comparisons of XCH4 time series from January 2011 to December 2014 over the selected study regions. The simulated XCH4 data are sampled at the nearest model grid to the available GOSAT observations and at the satellite overpass time (∼ 13:00 LT) and then averaged over each study region. Observations are sparse or not available during the SW monsoon season in some of the regions due to limitations of GOSAT retrieval under cloud cover. The model captures the salient features of the seasonal cycles at very high statistical significance (correlation coefficients, r>0.8; except for northeastern India; Table 1). The high ACTM-GOSAT correlations for the low-/no-emission regions suggest that transport and chemistry are accurately modeled in ACTM. Although we do not have the statistically significant number of observations for the SW monsoon period, the observed high GOSAT XCH4 are generally well simulated by ACTM over most of the study regions. Based on these comparisons, we can assume that model simulations can be used to understand XCH4 variability over the Indian region. Though we showed only the paired GOSAT and ACTM data that matched in time and location in Fig. 2b–k, we also confirmed that the correlation is high (r∼0.9) between the monthly averaged time series of GOSAT and ACTM averaged for the 4 years (2011–2014) when ACTM is not co-sampled at the GOSAT sampling points (Fig. S1 in the Supplement). These high correlations ensure representativeness of the data shown in Fig. 2b–k. Thus, the seasonal evolution of XCH4 using the ACTM simulations alone is expected to be fairly valid for different altitude layers (refer to Patra et al., 2011b, for comparison at the aircraft cruising altitude). Though the model is only validated for XCH4 in this study, comparisons with surface and independent aircraft CH4 observations have been shown in Patra et al. (2016).

As mentioned earlier, the persistent deep convection and mean circulation during the SW monsoon season significantly enhance CH4 in the upper troposphere (e.g., Xiong et al., 2009; Baker et al., 2012), coinciding with the period of high surface CH4 emissions due to rice paddy cultivation and wetlands over the Indian region (Yan et al., 2009; Hayashida et al., 2013). Although both these emissions and transport processes contribute greatly to seasonal changes in XCH4, their relative contributions have not been studied over the monsoon-dominated Indian region.

To understand the role of transport, the atmospheric column is segregated into five sigma-pressure (σp) layers, starting from the surface level (σp=1) to the top of the atmosphere (σp=0), with an equal layer thickness of σp=0.2. Lower Troposphere (LT), Mid-Troposphere1 (MT1), Mid-Troposphere2 (MT2), Upper Troposphere (UT), and Upper Atmosphere (UA) denote the layers corresponding to the sigma-pressure values of 1.0–0.8, 0.8–0.6, 0.6–0.4, 0.4–0.2, and 0.2–0.0. The partial columnar CH4 is calculated within different σp layers (denoted by XpCH4) using the same formula for XCH4, as in Sect. 2.2. The model results are averaged over each sub-region of our analysis for the XCH4 seasonal cycle. To understand the role of surface emission in the XCH4 seasonal cycle, the climatology of the optimized total CH4 flux for each sub-region is compared. Figure 3 shows the monthly mean climatology (average for 2011–2014) of total CH4 flux, XCH4, and XpCH4 from the model averaged over three selected regions, EIGP (a1–a7), SP (b1–b7), and AI (c1–c7). These representative regions have been selected because they show distinct XCH4 seasonal cycles and the dominant controlling factors (such as emission, transport, and chemistry). The observed GOSAT XCH4 values are also shown for a reference; however, the model results do not correspond to the location and time of GOSAT observations (as opposed to those in Fig. 2). The plots for the remaining seven regions are available in Figs. S2 and S3.

Over the EIGP region, the magnitude and timing of the seasonal peak in emission differ substantially between the CTL and AGS emission scenarios (refer to Fig. 3a7). ACTM simulated XCH4 seasonal peak is in agreement with the peak in emission in June for the AGS case (Fig. 3a6). However, simulated XCH4 remains nearly constant until September, although the emission decreases substantially toward winter. In general, the emission is relatively higher in the monsoon season (July–August–September) than in other seasons in both cases. However, in the LT, where we expect most susceptibility to the surface emission, the partial column CH4 indicates very different seasonality from the emissions; XpCH4 (LT) increases toward winter continuously (Fig. 3a5). The partial CH4 columns for the upper troposphere and middle troposphere (Fig. 3a2–a3) show similar seasonality to the total XCH4 rather than in the LT. Therefore, this analysis strongly suggests that the emissions from the surface and the upper tropospheric partial column both contribute to the formation of the XCH4 seasonal cycle. These results also suggest the possibility that GOSAT and ACTM XCH4 data can be used for correcting a priori emission scenarios by inverse modeling.

In contrast to the XCH4 seasonal cycle over EIGP, a notable difference is observed in the emission and XCH4 seasonal cycle over the SP region (Fig. 3b). The XCH4 seasonal cycle and emission seasonal cycle are found to be out of phase with each other and the differences in emission scenarios are not reflected in XCH4 seasonal variations. Both emission scenarios show the distinct seasonal pattern: AGS shows annual high emissions from April to September, while CTL shows an annual high during August–September (Fig. 3b7). The total emissions over SP are much lower than that of EIGP (note the different y axis scale for Fig. 3b7) and hence the difference between the XCH4 simulations from both emission scenarios is comparatively low. The XCH4 shows almost identical seasonal cycles for both of the emission scenarios, a peak in October, and prolonged low values during May–September. The seasonal XpCH4 cycle in the LT layer shows the seasonal pattern similar to the total XCH4. Inconsistency between emission seasonality and XCH4 coupled with low emissions strongly suggests that the XCH4 can be controlled by transport and/or chemistry but not emissions. Surface winds during May–September over SP are of marine origin, which effectively flushes the air with low CH4 (see Fig. S4). Further, the distinct seasonal cycle of chemical loss is observed over the SP region compared to other study regions; the loss rate starts increasing from 6 ppbday-1 in January to 12 ppbday-1 in April, and continues to remain high until September (refer to Fig. S5). These pieces of evidence clearly suggest that the combined effect of transport and chemistry causes the low XCH4 values for the May–September period over the SP region. The peaks in the upper layers in October (Fig. 3b1–b4) and transport from the polluted continental layer in the LT layer (refer to Fig. S4) could together contribute to the seasonal XCH4 peak over SP. Based on these findings, we conclude that the XCH4 measurements do not impose a strong constraint on surface emissions for inverse modeling over the SP region, suggesting a need for in situ measurements.

Over the Arid India (AI) region, the XCH4 seasonal cycle is observed to be different from those of the EIGP and SI regions. The simulated XCH4 (Fig. 3c6) shows extremely weak sensitivity to the surface emission differences between the AGS and CTL cases (Fig. 3c7). Additionally, the XpCH4 in the LT layer (Fig. 3c5) does not resemble the phase of seasonality in surface emissions and simulated/observed XCH4. The XpCH4 in the LT layer decreases from January to August and increases until December. On the other hand, a remarkable peak (∼ 1896 ppb) is observed in XCH4 during August, followed by a decline afterward (Fig. 3c6). This is an outstanding example of deceiving linkage between surface emissions and XCH4 in terms of seasonal variation. An enhancement in the mixing ratios of XpCH4 is observed from May to August only in the MT2 and UT layers (Fig. 3c2–c3) and from June to August in the UA layer (Fig. 3c1). This analysis infers that MT2 and UT partial columns mostly contribute to the formation of the XCH4 seasonal cycle over the AI region.

Next, we quantify the contributions of different partial layers (XpCH4) in the formation of XCH4 seasonal amplitude (Fig. 4). As the phase of the XpCH4 seasonal cycle does not always match with that of XCH4, we have fixed months of peak and trough in the XCH4 seasonal cycle for this analysis. First, we calculate the differences of the XpCH4 values at the time of the peak and the trough of the XCH4 over each region, and then the differences at different partial layers are divided by the seasonal amplitude of XCH4 for calculating the contributions from the respective layers into the seasonal amplitude of XCH4.

Figure 4 reveals that ∼ 40 % of the seasonal enhancement in the observed XCH4 can be attributed to the partial pressure layers below 600 hPa (LT and MT1) for the EIGP region, which is directly influenced by the surface emissions. About 40 % in seasonal enhancement comes from layers above 600 hPa. Over the SP region, about 60 % of the seasonal XCH4 amplitude is attributed to layers below 600 hPa and the remaining 40 % results from the upper layers. Although the activities in the lower atmosphere (below 600 hPa) govern most of the seasonal XCH4 cycle over this region, there is no clear link with seasonal variations in emissions as this region is under greater influence of changes in monsoon meteorology. These regions are under the influence of emission signals from the Indian subcontinent during winter, while in the summer, clean marine air controls CH4 levels (see also Patra et al., 2009). In contrast to the two regions mentioned above, over the AI region, the LT and MT1 layers together contribute only about 12 % to the formation of the XCH4 seasonal cycle amplitude, and the layers above 600 hPa contribute the remaining 88 %. These findings lead us to conclude that instead of surface emissions, the high CH4 in the upper tropospheric layers contributes significantly to the formation of seasonal peaks in XCH4.

The reason for high mixing ratios in the upper troposphere, as discussed in the previous section, can be explained by vertical transport of high CH4 emission signals from the surface, because the vertical transport timescales in the tropical region are much shorter than the chemical lifetime of CH4, on the order of 1–2 years (Patra et al., 2009). Figure 5a1–a4 show the latitude–pressure cross sections of the convective transport rate (in ppb day-1) and vertical velocity (hPas-1) averaged over 83–93∘ E for the different seasons of 2011 (the ACTM AGS case). The positive/negative values of the convective transport rate and vertical velocity in Fig. 5a1–a4 indicate the gain/loss of mass and downward/upward motions, respectively. Rapid updrafts of CH4, as indicated by higher negative vertical velocity, by deep convection during the monsoon season are aided by the regional topography of the IGP region (north of 20∘ N and east of 79∘ E in the Indian region). These updrafts lift CH4-rich air into the upper tropospheric region (Fig. 5b3). The CH4 concentrations at the surface level decreased rapidly at an average rate of ∼ 10 ppbday-1 during the SW monsoon season, and accumulate in the upper troposphere at a similar rate over the IGP region (Fig. 5a3). During the winter, spring, and autumn seasons surface CH4 decreased at an average rate of 2, 8, and 7 ppbday-1, respectively. CH4 levels accumulate in the middle and upper troposphere at an average rate of 6 ppbday-1 during the spring and autumn seasons, while during the winter season no significant accumulation has been observed at this height over the IGP region (Fig. 5a1, a2, and a4). Overall these transport processes repeat every year with a certain degree of interannual variation, as can be seen for the years from 2011 to 2014. The interannual variations are likely to have been caused by the early/late onset and retreat of the SW monsoon as well as the weak/strong monsoon activity over the years.

The horizontal cross sections of CH4 at 200 hPa are shown with wind vectors in Fig. 5c1–c4 for understanding the spatial extent of uplifted CH4-rich air over the whole South Asian region. The uplifted CH4-rich air mass is trapped in the upper troposphere (∼ 200 hPa) when encountered by the anticyclonic winds during the SW monsoon season. This leads to a widespread CH4 enhancement covering a large part of South Asia, and the CH4-rich air leaked predominantly along the southern side of the sub-tropical westerly jet over to East Asia (Fig. 5c3; see also Umezawa et al., 2012). As a result of this, the high CH4 air masses at the upper troposphere are not limited to the regions of intense surface emissions as discussed earlier. After the SW monsoon season, the strong westerly jet breaks the upper tropospheric anticyclone and the CH4-rich air mass shifts over southern India during the autumn season (Fig. 5c4). In this way, the convective updraft of high-CH4 air mass, followed by horizontal spreading of the air mass over the larger area by anticyclonic circulation, controls the redistribution of CH4 in the upper troposphere over the northern part of India during the SW monsoon season, and over the Southern Peninsula during the early autumn season.