Introduction

CH4 is the second most important anthropogenic greenhouse gas after CO2, accounting for 20% of direct anthropogenic radiative forcing1. CH4 contributes strongly to anthropogenic climate change, directly through its radiative forcing as well as indirectly through impacts on atmospheric chemistry2. With a relatively short atmospheric lifetime of ∼9 years, CH4 is a primary target for global warming mitigation strategies3. Over the past decades, the atmospheric CH4 growth rate has been highly variable3,4,5,6,7, including an approximate stabilization from 1999 to 2006 followed by a renewed growth since 20078. Among a range of explanations that were proposed, some studies have suggested that more than 70% of the interannual variations of CH4 can be explained by wetland CH4 emissions9,10.

Wetland CH4 emissions are highly sensitive to soil temperature and moisture11. Paleo records and studies of contemporary CH4 suggest a strong positive feedback of wetlands to global warming through CH4 emissions12,13. Proper quantification of this feedback is important for accurate future climate projections. Therefore, it is crucial to better understand the sensitivity of wetland CH4 emissions to changes in climatic parameters. The El Niño Southern Oscillation (ENSO) is a major mode of variability of global precipitation and temperature, comprising alternating El Niño and La Niña phases14. Hodson et al.15 estimated the influence of precipitation and temperature change, driven by ENSO, on wetland CH4 emissions using a process-based wetland model. They found that a large fraction of CH4 variability is correlated with ENSO, with higher tropical wetland CH4 emission during La Niña periods. However, this pattern has not been verified until now by atmospheric CH4 measurements during a La Niña. Furthermore, La Niña periods have received less attention in studies of the atmospheric CH4 budget than El Niño, since continued warming likely favors neutral or El Niño conditions16.

The La Niño of 2011 (LN11 hereafter) was the strongest since 1980 (see Fig. 1a) and offers the possibility to investigate the response of the atmospheric CH4 budget to La Niña conditions. In this study, we investigate this response by combining different measurement dataset and model simulations. A brief overview of them is given in the next section.

Figure 1
figure 1

(a) Multivariate ENSO index (MEI53). The strong La Niña of 2011 (LN11) is shaded in dark green. The preceding El Niño of 2010 (EN10) and succeeding weak La Niña of 2012 (LN12) are shaded in lighter red and green colors, respectively. (b,c,d) Detrended and smoothened XCH4 integrated over the large regions: (b) GOSAT FP XCH4, (c) TM5-Meteorology XCH4—that is, XCH4 variability due to meteorological changes (TM5 is run with annually repeating emissions). (d) GOSAT FP XCH4 corrected for the influence of meteorology (the difference between b and c). The light shaded regions represent the ±1σ uncertainty of the respective time series. NET, SET, TRO, and GLO are abbreviation of Northern Extra Tropics, Southern Extra Tropics, Tropics and Globe, respectively.

Full size image

Method and Data

Atmospheric CH4 measurements are available during the 2011 La Niña period from ground-based networks (NOAA-ESRL, CSIRO), and space (GOSAT, SCIAMACHY). The Greenhouse Gases Observing Satellite (GOSAT) has been measuring spectra for retrieval of the column average mole fraction of CH4 (XCH4) since June 200917. Onboard GOSAT is the Thermal And Near infrared Sensor for carbon Observation-Fourier Transform Spectrometer (TANSO-FTS), from which XCH4 is obtained with high sensitivity to the lower troposphere, and hence, to surface emissions18. We analyze the interannual variability in GOSAT full-physics (FP) XCH4, obtained using the RemoteC algorithm19, and ground-based CH4 flask-air measurements. Supplementary Material (SM) Section 9 further explains the FP retrieval method and justifies our choice of FP XCH4 over XCH4 derived from other retrieval algorithms.

In addition to the surface emissions, changes in atmospheric transport can cause interannual variability in CH420. Large-scale transport patterns, including the strength of inter-hemispheric exchange and atmospheric temperature are influenced by ENSO3,21,22,23. To quantify the contribution of these meteorological parameters, we ran the Tracer Transport Model version 5 (TM524) repeating surface emissions of 2008 for every year in 2009–2015. This simulation is referred to as TM5-Meteorology from hereon (see SM Section 1). To quantify the contribution of the surface emissions to XCH4 variability, we look at the difference between GOSAT FP XCH4 and XCH4 sampled from TM5-Meteorology.

Atmospheric inverse modeling systems are well established tools to convert atmospheric CH4 measurements into surface emissions25,26. We use the TM5-4DVAR (TM5-variational data assimilation system27) in combination with GOSAT FP XCH4, and surface measurements from NOAA-ESRL3 and CSIRO28 to optimize surface CH4 emissions. Note that the inverse model makes use of actual meteorological fields from the ECMWF ERA-interim reanalysis to account for variability in the atmospheric transport of CH4. Earlier studies have established the link between biomass burning CH4 emissions and ENSO29. To exclude the influence of biomass burning, fire related CH4 emissions from the Global Fire Emissions Database version 4s (GFED4s) inventory have been subtracted from the TM5-4DVAR emissions.

The origin of an atmospheric CH4 anomaly can be identified using CH4 stable isotope measurements. We look at measurements of 13C/12C in CH4 (expressed in δ-notation as δ13C-CH4) analyzed by INSTAAR in samples from the NOAA-ESRL (ref. 30, see Fig. 2b). δ13C-CH4 of atmospheric CH4 (global average in 2009=−47.14‰) is controlled by the relative contribution from different source types with distinct isotopic signatures. The mean isotopic signatures of the biogenic category is ∼ −60‰ (includes wetlands, agriculture, waste), for the thermogenic category it is ∼ −37‰ (includes fossil-fuels) and for pyrogenic category it is ∼ −22‰ (includes biomass burning)31. To account for impact of meteorological variability on δ13C-CH4, a meteorology simulation of δ13C-CH4 was performed using TM532.

Figure 2
figure 2

(a) Detrended and smoothened CH4 surface emissions estimates from TM5-4DVAR for the same regions as in Fig. 1. The variability of GFED4s biomass burning emissions has been subtracted. (b) δ13C-CH4 measurements54 corrected for the influence of transport using a meteorology-only TM5 simulation of δ13C-CH432. The light shaded regions represent the ±1σ uncertainty of the respective time series.

Full size image

To identify factors which might have altered the wetland emissions, we look at the variability in land precipitation and temperature data in CRU-TS version 3.23 (Climatic Research Unit-time series33). We also analyze CH4 emission and surface inundation extent from two process-based wetland models: LPJ-wsl15,34 and CLM4.535,36. Additionally, we derive an independent estimate of inundation extent from remotely sensed Surface WAter Microwave Product Series (SWAMPs37).

The primary sink of CH4 is the reaction with OH in the troposphere (∼454–617 TgCH4 yr−1 2), and interannual variations in OH can also contribute to the observed CH4 variability. Tropospheric OH concentrations are influenced by many factors, including temperature, water vapor, O3, NOx, CH4, CO, and the overhead stratospheric ozone column38. The TM5-4DVAR inversions performed in this study make use of OH fields from ref. 39, which vary seasonally, but are the same each year. To investigate possible variations caused by the OH sink, we analyze posterior CH4 emissions of LMDz-PYVAR-SACS inversion40,41,42. In this inversion, the OH fields were optimized simultaneously using methyl chloroform (MCF) measurements (see SM Section 1.3).

Data Analysis

The above mentioned measurements and model outputs have been analyzed by taking their monthly averages and integrating them over three zones over the globe (GLO) : Tropics (TRO: 30°S to 30°N), Northern Extra Tropics (NET: 30°N to 90°N), Southern Extra Tropics (SET: 90°S to 30°S). The time series of these monthly averages have been detrended and smoothened using a 12 month running mean. The resulting time series of GOSAT FP XCH4 from this method is less influenced by systematic errors, which often affect the satellite retrievals9,43,44.

Further, based on the Multivariate ENSO index (MEI) we define three periods: 1) La Niña in 2011 as LN11; 2) The preceding El Niño of 2010 as EN10; 3) succeeding weak La Niña of 2012 as LN12 (see Fig. 1a). Time series of measurements and model outputs are analyzed during these periods using the following parameters:

  1. 1

    : Defined as the change of a quantity q in region r during an ENSO phase.

  2. 2

    : Defined as the mean of a quantity q in region r during an ENSO phase.

Table 1 summarizes or of time series shown in Figs 1 and 2.

Table 1 (i.e. the sum of the derivative) or (i.e. mean) of the times series of quantity q, averaged over region r, as shown in Figs 1 and 2 during different ENSO phases.
Full size table

Results and Disscussion

GOSAT observations

Figure 1b shows detrended and smoothened time series of GOSAT FP XCH4. During LN11, SET and TRO have increasing XCH4 ( = 5.6ppb and  = 2.4) contrasted by a decrease in NET ( = −2.7 ppb). During LN12, the opposite trends are found: XCH4 gradually reduced in SET and TRO and increased in NET.

Figure 1c shows the results of TM5-Meteorology simulation. The modeled XCH4 decreases over NET and increases over SET ( ppb, ppb). These trends can be attributed to the faster inter-hemispheric exchange during La Niña conditions, which transferred additional CH4 rich air from the Northern Hemisphere to the Southern Hemisphere. Francey et al.21 found the strongest inter-hemispheric transport of CH4 during LN11 since 1990. Previous studies have reported such enhancement in inter-hemispheric transport, and consequent increase in CH4 in the Southern Hemisphere, during the La Niña of 2007–20083 and 198923. During LN12, the strength of anomalies over both regions weakens as inter-hemispheric exchange returns to its normal strength. Modeled XCH4 over TRO show less variation during LN11. During LN12, the modeled XCH4 ( ppb) explains a major fraction of GOSAT XCH4 ( ppb) variability. It is noteworthy that OH concentrations are not affected in our meteorology simulation as TM5 uses OH fields from Spivakovsky et al.39. Eventhough these fields don’t vary interannually, temperature variations can still cause variability in the atmospheric CH4 sink, as the rate constant for reaction with OH is temperature dependent45.

Figure 1d shows the GOSAT FP XCH4 after correction for variations in atmospheric transport using the TM5-Meteorology simulation, and indicates the fraction of CH4 variability that can be attributed to variability in CH4 sources and sinks. An analogous plot showing surface flask-air measurements is given in the SM section 4 and shows similar patterns. During LN11, the largest transport-corrected is seen in TRO ( ppb), pointing to higher CH4 emissions from the tropical continents. As SET does not have large CH4 surface emissions, it is caused most likely by transport from TRO. This means the transport-corrected anomaly of TRO is transferred to SET.

Emissions and source attribution

Detrended and smoothened time series of the posterior TM5-4DVAR emissions are shown in Fig. 2a. The CH4 emissions over TRO show a positive anomaly during LN11 ( TgCH4 yr−1). is higher by 11.7 TgCH4 yr−1 in LN11 than in the preceding EN10 ( TgCH4 yr−1). This can be due to higher CH4 emissions from tropical wetlands during the La Niña, as suggested by ref. 15. During LN12, the CH4 emission enhancement over TRO is weaker ( TgCH4 yr−1). NET has a sharp decrease in CH4 emissions during LN11 ( TgCH4 yr−1), which shifts the maximum of the global CH4 emission anomaly towards the beginning of LN11.

Detrended and smoothened time series of δ13C-CH4 is shown in Fig. 2b. During LN11, δ13C-CH4 decreased over each region. Over TRO, we observe a decrease of 0.06‰, which is a larger decrease than the GLO δ13C-CH4 decrease by 0.03‰. An isotope mass balance calculation shows that if the increase in TRO CH4 emissions of 11.7 TgCH4 yr−1 (change from EN10 with TgCH4 yr−1 to LN11 with TgCH4 yr−1), is attributed to a biogenic source, it would cause a drop in δ13C-CH4 of similar magnitude. This indicates that the source of the LN11 CH4 anomaly in the Tropics is of biogenic origin. The increase of GLO δ13C-CH4 (≈0.03‰) over LN12 can be explained by reduced biogenic emissions ((LN12) - (LN11) = ∼ −14 TgCH4 yr−1) and increased biomass burning emissions ( (LN12) – (LN11) = 2.2 TgCH4 yr−1) in comparison to LN11.

Tropical Biomass burning is strongly influenced by ENSO46. Globally, GFED4s biomass burning CH4 emissions indicate a decrease of 0.67 TgCH4 yr−1 from EN10 to LN11 (see SM Section 5). The effect of this change on the isotopic composition is only −0.003‰, thus much smaller than the observed trend. According to GFED4s, CH4 emissions from biomass burning over TRO during LN11 are close to average over the whole period ( = −0.23 TgCH4 yr−1). In LN12, these emissions were higher in NET and TRO ( = 1.09 TgCH4 yr−1 and  = 1.32 TgCH4 yr−1). This increase may be explained by higher fuel availability due to enhanced biomass growth during the preceding LN11. Ref. 47 suggested a similar impact of Australian biomass burning on CO2 emissions.

Figure 3a and b show monthly anomalies recorded in climate parameters. A significant redistribution of heat and precipitation is seen during the different phases of ENSO. was −1.72, 4.90, and 0.64 mm during EN10, LN11, and LN12, respectively. During LN11 the precipitation anomaly in TRO (and in GLO) was the highest since the onset of the 21st century (see SM Figure 10). Regions like Australia had six consecutive seasons of increased rainfall over the La Niña of 2011 and 201248. Higher temperatures were observed in NET during LN11 ( = 0.18 °C), favoring increased biomass burning, for example, near Moscow during the summer of 201049,50. Mean temperatures during LN11 ( = −0.05 °C) were in between those during EN10 ( = 0.15 °C) and LN12 ( = −0.22 °C).

Figure 3
figure 3

(a,b) Detrended and smoothened regionally averaged precipitation and temperature measurements over land in CRU-TS version 3.23 (Climatic Research Unit-time series33). (c) Anomalies in the total inundated area estimated by SWAMPS37.

Full size image

An increase in total inundated area is observed in the remotely sensed SWAMPS data (see Fig. 3c). The total inundated area estimated by the wetland models LPJ-wsl and CLM4.5 also show a similar increase. However, these wetland models estimate a relatively weaker enhancement in CH4 emissions with TgCH4 yr−1 for LPJ-wsl and TgCH4 yr−1 for CLM4.5 during LN11 (see SM Section 8).

To further investigate the relation between inversion-estimated CH4 emissions and potential climatic drivers, we examine their correlation coefficients (R) [see SM Figure 11]. CH4 emission anomalies (as shown in Fig. 2a) correlate stronger with precipitation anomalies than with temperature anomalies (as shown in Fig. 3) in both NET , ) and TRO (, ). This points to precipitation as the more important driver of the CH4 anomaly in TRO during LN11, supported further by the correlation with inundated area (). This is consistent with the findings of Bloom et al.51, who show that precipitation plays a more dominant role than temperature in determining anomalous CH4 variability in the Tropics.

To investigate possible variations caused by the OH sink, we analyzed optimized CH4 emissions from a LMDz-PYVAR-SACS inversion, in which OH fields were also optimized. During LN11, the results of this inversion suggest of 9.1 TgCH4 yr−1, compared to TM5-4DVAR of ∼6 TgCH4 yr−1 (see SM Section 6). The differences in interannual variations of the emission estimates of the two inversions are mainly caused by their different treatment of OH sink. Assuming that the MCF-optimized OH sink of LMDz-PYVAR-SACS is more accurate than TM5, the sink was stronger than normal during LN11 by ∼3 TgCH4 yr−1. This is consistent with the hypothesis of an increased CH4 sink during La Niña and a weaker sink during El Niño52.

Conclusion

Our inversion results, supported by δ13C-CH4 measurements, provide strong evidence of enhanced tropical biogenic CH4 emissions by ∼6–9 TgCH4 yr−1, during the La Niña of 2011. Wetlands were the likely cause of this anomaly as a simultaneous increase in total inundated area is shown by remote sensing observations and hydrological models. The increase in inundated area was in response to La Niña induced increase in precipitation. 2011 experienced the strongest La Niña event in the past 4 decades as well as since the onset of modern atmospheric CH4 measurements. It is noteworthy that during this La Niña the increase in global CH4 mole fractions were not as pronounced due to a simultaneous decrease in the CH4 emissions in the Northern Extra Tropics. Our analysis presents the first evidence of the large-scale response of wetland CH4 emissions to ENSO variability using satellite retrievals.

Data Availability

We use Level 2 SRFP XCH4 v2.3.7 GOSAT XCH4 retrievals that are publicly available from ESA’s Climate Change Initiative website (www.esa-ghg-cci.org/). NOAA CH4 and INSTAAR δ13C-CH4 measurements are freely available from NOAA’s public ftp server (ftp://aftp.cmdl.noaa.gov/data). CSIRO CH4 measurements can be downloaded from the WDCGG (World Data Centre for Green-house Gases) website. GFED4s CH4 emissions can be downloaded from http://daac.ornl.gov. CRU TS3.23). Precipitation and temperature data are held at British Atmospheric Data Centre, RAL, UK (http://badc.nerc.ac.uk/data/cru/). SWAMP wetlands fraction data can be downloaded from http://wetlands.jpl.nasa.gov after a short registration. CLM4.5 and LPJ-wsl CH4 emissions and wetlands fractions can be obtained by contacting William J. Riley and B. Poulter, respectively.

Additional Information

How to cite this article: Pandey, S. et al. Enhanced methane emissions from tropical wetlands during the 2011 La Niña. Sci. Rep. 7, 45759; doi: 10.1038/srep45759 (2017).

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.