What controls the seasonal cycle of columnar methane observed by GOSAT over different regions in India?

Methane (CH4) is one of the most important shortlived climate forcers for its critical roles in greenhouse warming and air pollution chemistry in the troposphere, and the water vapor budget in the stratosphere. It is estimated that up to about 8 % of global CH4 emissions occur from South Asia, covering less than 1 % of the global land. With the availability of satellite observations from space, variability in CH4 has been captured for most parts of the global land with major emissions, which were otherwise not covered by the surface observation network. The satellite observation of the columnar dry-air mole fractions of methane (XCH4) is an integrated measure of CH4 densities at all altitudes from the surface to the top of the atmosphere. Here, we present an analysis of XCH4 variability over different parts of India and the surrounding cleaner oceanic regions as measured by the Greenhouse gases Observation SATellite (GOSAT) and simulated by an atmospheric chemistrytransport model (ACTM). Distinct seasonal variations of XCH4 have been observed over the northern (north of 15 N) and southern (south of 15 N) parts of India, corresponding to the peak during the southwestern monsoon (July–September) and early autumn (October–December) seasons, respectively. Analysis of the transport, emission, and chemistry contributions to XCH4 using ACTM suggests that a distinct XCH4 seasonal cycle over northern and southern regions of India is governed by both the heterogeneous distributions of surface emissions and a contribution of the partial CH4 column in the upper troposphere. Over most of the northern Indian Gangetic Plain regions, up to 40 % of the peak-to-trough amplitude during the southwestern (SW) monsoon season is attributed to the lower troposphere (∼ 1000–600 hPa), and ∼ 40 % to uplifted high-CH4 air masses in the upper troposphere (∼ 600–200 hPa). In contrast, the XCH4 seasonal enhancement over semi-arid western India is attributed mainly (∼ 70 %) to the upper troposphere. The lower tropospheric region contributes up to 60 % in the XCH4 seasonal enhancement over the Southern Peninsula and oceanic region. These differences arise due to the complex atmospheric transport mechanisms caused by the seasonally varying monsoon. The CH4 enriched air mass is uplifted from a high-emission region of the Gangetic Plain by the SW monsoon circulation and deep cumulus convection and then confined by anticyclonic wind in the upper tropospheric heights (∼ 200 hPa). The anticyclonic confinement of surface emission over a wider South Asia region leads to a strong contribution of the upper troposphere in the formation of the XCH4 peak over northern India, including the semi-arid regions with extremely low CH4 emissions. Based on this analysis, we suggest that a link between surface emissions and higher levels of XCH4 is not always valid over Asian monsoon regions, although there is often a fair correlation between surface emissions and XCH4. The overall validity of ACTM simulation for capturing GOSAT observed seasonal and spatialXCH4 variability will allow us to perform inverse modeling ofXCH4 emissions in the future usingXCH4 data. Published by Copernicus Publications on behalf of the European Geosciences Union. 12634 N. Chandra et al.: Controlling factors for XCH4 seasonality over India

Abstract.Methane (CH 4 ) is one of the most important shortlived climate forcers for its critical roles in greenhouse warming and air pollution chemistry in the troposphere, and the water vapor budget in the stratosphere.It is estimated that up to about 8 % of global CH 4 emissions occur from South Asia, covering less than 1 % of the global land.With the availability of satellite observations from space, variability in CH 4 has been captured for most parts of the global land with major emissions, which were otherwise not covered by the surface observation network.The satellite observation of the columnar dry-air mole fractions of methane (XCH 4 ) is an integrated measure of CH 4 densities at all altitudes from the surface to the top of the atmosphere.Here, we present an analysis of XCH 4 variability over different parts of India and the surrounding cleaner oceanic regions as measured by the Greenhouse gases Observation SATellite (GOSAT) and simulated by an atmospheric chemistrytransport model (ACTM).Distinct seasonal variations of XCH 4 have been observed over the northern (north of 15 • N) and southern (south of 15 • N) parts of India, corresponding to the peak during the southwestern monsoon (July-September) and early autumn (October-December) seasons, respectively.Analysis of the transport, emission, and chemistry contributions to XCH 4 using ACTM suggests that a distinct XCH 4 seasonal cycle over northern and southern regions of India is governed by both the heterogeneous distributions of surface emissions and a contribution of the partial CH 4 column in the upper troposphere.Over most of the northern Indian Gangetic Plain regions, up to 40 % of the peak-to-trough amplitude during the southwestern (SW) monsoon season is attributed to the lower troposphere (∼ 1000-600 hPa), and ∼ 40 % to uplifted high-CH 4 air masses in the upper troposphere .In contrast, the XCH 4 seasonal enhancement over semi-arid western India is attributed mainly (∼ 70 %) to the upper troposphere.The lower tropospheric region contributes up to 60 % in the XCH 4 seasonal enhancement over the Southern Peninsula and oceanic region.These differences arise due to the complex atmospheric transport mechanisms caused by the seasonally varying monsoon.The CH 4 enriched air mass is uplifted from a high-emission region of the Gangetic Plain by the SW monsoon circulation and deep cumulus convection and then confined by anticyclonic wind in the upper tropospheric heights (∼ 200 hPa).The anticyclonic confinement of surface emission over a wider South Asia region leads to a strong contribution of the upper troposphere in the formation of the XCH 4 peak over northern India, including the semi-arid regions with extremely low CH 4 emissions.Based on this analysis, we suggest that a link between surface emissions and higher levels of XCH 4 is not always valid over Asian monsoon regions, although there is often a fair correlation between surface emissions and XCH 4 .The overall validity of ACTM simulation for capturing GOSAT observed seasonal and spatial XCH 4 variability will allow us to perform inverse modeling of XCH 4 emissions in the future using XCH 4 data.

Introduction
Methane (CH 4 ) is the second most important anthropogenic greenhouse gas (GHG) after carbon dioxide (CO 2 ) and accounts for ∼ 20 % (+0.97 W m −2 ) of the increase in total direct radiative forcing since 1750 (Myhre et al., 2013).CH 4 is emitted from a range of anthropogenic and natural sources on the Earth's surface into the atmosphere.The main natural sources of CH 4 include wetlands and termites (Matthews and Fung, 1987;Cao et al., 1998;Sugimoto et al., 1998).Livestock, rice cultivation, the fossil fuel industry (production and uses of natural gas, oil, and coal), and landfills are the major sectors among the anthropogenic sources (Crutzen et al., 1986;Minami and Neue, 1994;Olivier et al., 2005;Yan et al., 2009).These results also suggest that the Asian region is an emission hotspot of CH 4 due to the large number of livestock, intense cultivation, coal mining, waste management, and other anthropogenic activities (EDGAR2FT, 2013).
With a short atmospheric lifetime of about 10 years (e.g., Patra et al., 2011a) and having 34 times more potential to trap heat than CO 2 on a mass basis over a 100-year timescale (Gillett and Matthews, 2010;Myhre et al., 2013), mitigation of CH 4 emissions could be the most important way to limit global warming at inter-decadal timescales (Shindell et al., 2009).Better knowledge of CH 4 distribution and quantification of its emission flux is indispensable for assessing possible mitigation strategies.However, sources of CH 4 are not yet well quantified due to sparse ground-based measurements, which results in limited representation of CH 4 flux on a larger scale (Dlugokencky et al., 2011;Patra et al., 2016).Recent technological advances have made it possible to detect spatial and temporal variations in atmospheric CH 4 from space (Frankenberg et al., 2008;Kuze et al., 2009), which could fill the gaps left by ground-, aircraft-and shipbased measurements, albeit at a lower accuracy than the in situ measurements.Further, despite the satellite observations having an advantage of providing continuous monitoring over a wide spatial range, the information obtained from passive nadir sensors that use solar radiation at the short-wavelength infrared (SWIR) spectral band is limited to columnar dry-air mole fractions of methane (XCH 4 ).This is an integrated measure of CH 4 with contributions from the different vertical atmospheric layers, i.e., from the measurement point on the Earth's surface to the top of the atmosphere (up to about 100 km or more precisely to the satellite orbit).
The South Asia region, consisting of India, Pakistan, Bangladesh, Nepal, Bhutan, and Sri Lanka, exerts a significant impact on the global CH 4 emissions, with regional total emissions of 37 ± 3.7 Tg CH 4 of about 500 Tg CH 4 global total emissions during the 2000s (Patra et al., 2013).The Indo-Gangetic Plain (IGP) located in the foothills of the Himalayas is one of the most polluted regions in the world, hosts 70 % of coal-fired thermal power plants in India, and experiences intense agricultural activity (Kar et al., 2010).This region is of particular interest mainly due to the coexistence of deep convection and large emission of pollutants (including CH 4 ) from a variety of natural and anthropogenic sources.Rainfall during the SW monsoon season causes higher CH 4 emissions from the paddy fields and wetlands (e.g., Matthews and Fung, 1987;Yan et al., 2009;Hayashida et al., 2013), while the persistent deep convection results in updraft of CH 4 -laden air mass from the surface to the upper troposphere during the same season, which is then confined by anticyclonic winds at this height (Patra et al., 2011b;Baker et al., 2012;Schuck et al., 2012).Several other studies have also highlighted the role of convective transport of pollutants (including CH 4 ) from the surface to the upper troposphere (400-200 hPa) during the SW monsoon season (July-September) (Park et al., 2004;Randel et al., 2006;Xiong et al., 2009;Lal et al., 2014;Chandra et al., 2016).The dynamical system dominated by deep convection and anticyclone covers mostly the northern Indian region (north of 15 • N) due to the presence of the Himalayas and the Tibetan Plateau, while such a complex dynamical system has not been observed over the southern part of India (south of 15 • N) (Rao, 1976).
Satellite-based measurements show elevated levels of XCH 4 over the northern part of India (north of 15 • N) to be particularly high over the IGP during the SW monsoon season (July-September) and over southern India (south of 15 • N) during the early autumn season (October-December) (Frankenberg et al., 2008(Frankenberg et al., , 2011;;Hayashida et al., 2013).Previous studies have linked these high XCH 4 levels to the strong surface CH 4 emissions particularly from the rice cultivation over the Indian region because they showed statistically significant correlations over certain regions (Hayashida et al., 2013;Kavitha et al., 2016).The differences in the peak of the XCH 4 seasonal cycle over the northern and southern regions of India are also discussed on the basis of agricultural practice in India that takes place in two seasons, May-October and November-April, respectively.However, inferring local emissions directly from variations in XCH 4 is ambiguous, particularly over the Indian regions under the influence of monsoon meteorology, because XCH 4 involves contributions of CH 4 abundances from all altitudes along the solar light path.
This study attempts for the first time to separate the factors responsible (emission, transport, and chemistry) for the distributions of columnar methane (XCH 4 ) over the Asian monsoon region for different altitude segments.The XCH 4 mixing ratios are used for this study as observed from GOSAT and simulated by JAMSTEC's ACTM.We aim to understand relative contributions of surface emissions and transport in the formation of XCH 4 seasonal cycles over different parts of India and the surrounding oceans.This understanding will help us in developing an inverse modeling system for estimation of CH 4 surface emissions using XCH 4 observations and ACTM forward simulation.

Satellite data
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 dryair mole fractions of the two important greenhouse gases (XCH 4 and XCO 2 ) 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 XCH 4 data are available during the SW monsoon over South Asia.This study uses the GOSAT SWIR XCH 4 (Version 2.21)-Research Announcement product for the period of 2011-2014.The ground-based FTS measurements of XCH 4 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 simulations
Model analysis is comprised of simulations from JAM-STEC'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-CH 4 (Patra et al., 2011a) and used in inverse modeling of CH 4 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 CH 4 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: where CH 4 is the methane burden in the atmosphere, S CH 4 is the total emissions/sinks of CH 4 at the surface, L CH 4 is the total loss of CH 4 in the atmosphere due to the chemical reactions, and ∇φ is the transport of CH 4 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(Patra et al., , 2011a(Patra et al., , 2016)).
XCH 4 is calculated from the ACTM profile using the following equations: where CH 4 (n) is the dry-air mole fraction at model midpoint level, n = number of vertical sigma pressure layers of ACTM (= 1-60 with σ p values of 1.0 and 0.005), and σ p = 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 XCH 4 .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 XCH 4 .Because the focus of this study is seasonal and spatial variations in XCH 4 , a constant offset adjustment should not affect the main conclusions.
Optimized emissions are shown from a global inversion of surface CH 4 concentrations (Patra et al., 2016) and multiplied by a constant factor of 12 for a clear visualization.The ACTM is first sampled at the location and time of GOSAT observations and then seasonally averaged.The white spaces in first two columns (a1, a2, b1, b2) are due to the missing data caused by satellite retrieval limitations under cloud cover.
3 Results and discussion 3.1 XCH 4 over the Indian region: view from GOSAT and ACTM simulations This section presents an analysis of XCH 4 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 XCH 4 pattern in detail depending on the distinct spatial pattern of surface emissions and XCH 4 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 XCH 4 (Fig. 1a1-c2) and our knowledge of seasonal meteorological conditions.Since general features of XCH 4 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 XCH 4 mixing ratios are lower in spring and higher in autumn.A strong latitudinal gradient in XCH 4 is observed between the Indo-Gangetic Plain (IGP) and the other parts of India.XCH 4 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 XCH 4 , 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 CH 4 fluxes (AGS and CTL) show high emissions over the IGP and northeastern Indian regions (Fig. 1c1-c2).Most elevated levels of XCH 4 are often observed simultaneously with the higher emissions, suggesting a link between the enhanced XCH 4 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, XCH 4 is higher in autumn than in spring, though the emissions are higher in spring.
Figure 2b-k show ACTM-GOSAT comparisons of XCH 4 time series from January 2011 to December 2014 over the selected study regions.The simulated XCH 4 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 XCH 4 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 XCH 4 variability over the Indian region.Though we showed only the paired GOSAT and ACTM data that matched in time and location in Fig. 2bk, 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 XCH 4 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 XCH 4 in this study, comparisons with surface and independent aircraft CH 4 observations have been shown in Patra et al. (2016).

Seasonal cycle of XCH 4 and possible controlling factors
As mentioned earlier, the persistent deep convection and mean circulation during the SW monsoon season significantly enhance CH 4 in the upper troposphere (e.g., Xiong et al., 2009;Baker et al., 2012), coinciding with the period of high surface CH 4 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 XCH 4 , their relative contributions have not been studied over the monsoon-dominated Indian region.
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 CH 4 is calculated within different σ p layers (denoted by X p CH 4 ) using the same formula for XCH 4 , as in Sect.2.2.The model results are averaged over each sub-region of our analysis for the XCH 4 seasonal cycle.To understand the role of surface emission in the XCH 4 seasonal cycle, the climatology of the optimized total CH 4 flux for each sub-region is compared.Figure 3 shows the monthly mean climatology (average for 2011-2014) of total CH 4 flux, XCH 4 , and X p CH 4 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 XCH 4 seasonal cycles and the dominant controlling factors (such as emission, transport, and chemistry).The observed GOSAT XCH 4 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 XCH 4 seasonal peak is in agreement with the peak in emission in June for the AGS case (Fig. 3a6).However, simulated XCH 4 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 CH 4 indicates very different seasonality from the emissions; X p CH 4 (LT) increases toward winter continuously (Fig. 3a5).The partial CH 4 columns for the upper troposphere and middle troposphere (Fig. 3a2-a3) show similar seasonality to the total XCH 4 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 XCH 4 seasonal cycle.These results also suggest the possibility that GOSAT and ACTM XCH 4 data can be used for correcting a priori emission scenarios by inverse modeling.
In contrast to the XCH 4 seasonal cycle over EIGP, a notable difference is observed in the emission and XCH 4 seasonal cycle over the SP region (Fig. 3b).The XCH 4 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 XCH 4 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 XCH 4 simulations from both emission scenarios is comparatively low.The XCH 4 shows almost identical seasonal cycles for both of the emission scenarios, a peak in October, and prolonged low values during May-September.The seasonal X p CH 4 cycle in the LT layer shows the seasonal pattern similar to the total XCH 4 .Inconsistency between emission seasonality and XCH 4 coupled with low emissions strongly suggests that the XCH 4 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 CH 4 (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 ppb day −1 in January to 12 ppb day −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 XCH 4 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 XCH 4 peak over SP.Based on these findings, we conclude that the XCH 4 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 XCH 4 seasonal cycle is observed to be different from those of the EIGP and SI regions.The simulated XCH 4 (Fig. 3c6) shows extremely weak sensitivity to the surface emission differences between the AGS and CTL cases (Fig. 3c7).Additionally, the X p CH 4 in the LT layer (Fig. 3c5) does not resemble the phase of seasonality in surface emissions and simulated/observed XCH 4 .The X p CH 4 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 XCH 4 during August, followed by a decline afterward (Fig. 3c6).This is an outstanding example of deceiving linkage between surface emissions and XCH 4 in terms of seasonal variation.An enhancement in the mixing ratios of X p CH 4 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 XCH 4 seasonal cycle over the AI region.
Next, we quantify the contributions of different partial layers (X p CH 4 ) in the formation of XCH 4 seasonal amplitude (Fig. 4).As the phase of the X p CH 4 seasonal cycle does not always match with that of XCH 4 , we have fixed months of peak and trough in the XCH 4 seasonal cycle for this analysis.First, we calculate the differences of the X p CH 4 values at the time of the peak and the trough of the XCH 4 over each region, and then the differences at different partial layers are divided by the seasonal amplitude of XCH 4 for calculating the contributions from the respective layers into the seasonal amplitude of XCH 4 .
Figure 4 reveals that ∼ 40 % of the seasonal enhancement in the observed XCH 4 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 XCH 4 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 XCH 4 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 CH 4 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 XCH 4 seasonal cycle amplitude, and the layers above 600 hPa con-  Pressure (hPa) tribute the remaining 88 %.These findings lead us to conclude that instead of surface emissions, the high CH 4 in the upper tropospheric layers contributes significantly to the formation of seasonal peaks in XCH 4 .

Source of high CH 4 in the upper troposphere
The reason for high mixing ratios in the upper troposphere, as discussed in the previous section, can be explained by vertical transport of high CH 4 emission signals from the surface, because the vertical transport timescales in the tropical region are much shorter than the chemical lifetime of CH 4 , 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 (hPa s −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 CH 4 , 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 CH 4 -rich air into the upper tropospheric region (Fig. 5b3).The CH 4 concentrations at the surface level decreased rapidly at an average rate of ∼ 10 ppb day −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 CH 4 decreased at an average rate of 2, 8, and 7 ppb day −1 , respectively.CH 4 levels accumulate in the mid-dle and upper troposphere at an average rate of 6 ppb day −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 CH 4 at 200 hPa are shown with wind vectors in Fig. 5c1-c4 for understanding the spatial extent of uplifted CH 4 -rich air over the whole South Asian region.The uplifted CH 4 -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 CH 4 enhancement covering a large part of South Asia, and the CH 4 -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 CH 4 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 CH 4 -rich air mass shifts over southern India during the autumn season (Fig. 5c4).In this way, the convective updraft of high-CH 4 air mass, followed by horizontal spreading of the air mass over the larger area by anticyclonic circulation, controls the redistribution of CH 4 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.

Conclusions
The seasonal variations in dry-air mole fractions of methane (XCH 4 ) measured by the Greenhouse gases Observation SATellite (GOSAT) are analyzed over India and the surrounding seas using JAMSTEC's atmospheric chemistrytransport model (ACTM).The region of interest (the Indian landmass) is divided into eight sub-regions, namely, 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).The ACTM simulations are conducted using a couple of surface fluxes optimized by the inverse analysis as described in Patra et al. (2016).We have shown that the distinct spatial and temporal variations of XCH 4 observed by GOSAT are governed not only by the heterogeneity in surface emissions, but also by complex atmospheric transport mechanisms caused by the seasonally varying Asian monsoon.The seasonal XCH 4 patterns often show a fair correlation between emissions and XCH 4 over the regions resid-ing in the northern half of India (north of 15 • N: NEI, EI, EIGP, WIGP, CI, WI, AI), which would imply XCH 4 levels are closely associated with the distribution of emissions on the Earth's surface.However, detailed analysis of transport and emission using ACTM over these regions (except for AI) reveals that about 40 % of seasonal enhancement in the observed XCH 4 can be attributed to the lower tropospheric layer (below 600 hPa).The lower tropospheric layers are affected either by the surface emissions, e.g., in the northern India regions or seasonal changes in horizontal winds due to monsoon for the SP region.Up to 40 % of the seasonal CH 4 enhancement is found to come from the uplifted air mass into the 600-200 hPa height layer over northern regions in India.In contrast, over the semi-arid AI region, as much as ∼ 88 % of contributions to the XCH 4 seasonal cycle amplitude came from the height above 600 hPa, and only ∼ 12 % are contributed by the atmosphere below 600 hPa.The primary cause of the higher contributions from above 600 hPa over the northern Indian region is the characteristic of air mass transport mechanisms in the Asian monsoon region.The persistent deep convection during the southwestern monsoon season (June-August) causes strong updrafts of CH 4rich air mass from the surface to upper tropospheric heights (∼ 200 hPa), which is then confined by anticyclonic winds at this height.The anticyclonic confinement of surface emission over a wider South Asia region leads to strong contribution of the upper troposphere in formation of the XCH 4 peak over most regions in northern India, including the semi-arid regions with extremely low CH 4 emissions.In contrast to these regions, over the SP region, the major contributions (about 60 %) to XCH 4 seasonal amplitude come from the lower atmosphere (∼ 1000-600 hPa).Both transport and chemistry dominate in the lower troposphere over the SP region and thus the formation of the XCH 4 seasonal cycle is not consistent with the seasonal cycle of local emissions.As the upper level anticyclone does not cover the southern Indian region during the active phase of southwestern monsoon, no enhancement in XCH 4 is observed over the Southern Peninsula region.
This study shows that ACTM simulations are capturing the GOSAT observed seasonal and spatial XCH 4 variability well, and results provide an improved understanding of emissions, chemistry, and transport of CH 4 over one of the strongest global monsoonal regions.
Data availability.The satellite GOSAT records used in this study are available on the GOSAT official website (https://data2.gosat.nies.go.jp/index_en.html).The model simulation data could be available on request.The corresponding author and Prabir Patra (prabir@jamstec.go.jp) may be contacted for the same.

Figure 2 .
Figure 2. (a) The map of the regional divisions (shaded) for the time series analysis.(b-k) Time series of XCH 4 over the selected regions (shown on the map) as obtained from GOSAT and simulated by ACTM for two different emission scenarios, namely, ACTM_AGS and ACTM_CTL.The gaps are due to the missing observational data.

Figure 4 .
Figure 4. Contributions of partial columns in the seasonal amplitude of XCH 4 over selected regions for the AGS case.Differences in the X p CH 4 , calculated at the same time as the maxima and minima of the seasonal XCH 4 cycle, are used to calculate the percentage contributions of respective partial columns in the seasonal amplitude of XCH 4 .

Figure 5 .
Figure 5. Vertical structure of seasonally averaged CH 4 transport rate due to the convection (a1-a4, in ppb day −1 ) and CH 4 mixing ratios (b1-b4 from AGS scenarios) averaged over 83-93 • E for the year 2011.Positive and negative transport rate values represent the accumulation and dissipation of mass, respectively.The contour lines in the first (a1-a4) and second (b1-b4) columns depict the average omega velocity (in hPa s −1 ) and u wind component, respectively, for the same period.The solid contour lines show the positive values and the dotted lines show negative values.Positive and negative values of the omega velocity represent downward and upward motions, respectively.The zero value of u wind indicates that the wind is either purely southerly or northerly.White spaces in zonal-mean plots (a1-b4) show the missing data due to orography.The rightmost column (c1-c4) depicts the maps of averaged CH 4 and wind vectors (in m s −1 ; arrow) during all four seasons in 2011 at 200 hPa height.

Table 1 .
Correlation coefficients (r) between observed and model simulated seasonal cycles of XCH 4 .Model simulations are obtained from ACTM using two different emission scenarios, AGS and CTL.