Report

Large-Scale Controls of Methanogenesis Inferred from Methane and Gravity Spaceborne Data

See allHide authors and affiliations

Science  15 Jan 2010:
Vol. 327, Issue 5963, pp. 322-325
DOI: 10.1126/science.1175176

Measuring Methanogenesis

After carbon dioxide, methane is the second most important greenhouse gas, and an important species in terms of its role in atmospheric chemistry. The sources and sinks of methane, particularly the natural ones, are too poorly quantified, however, even to explain why the decades-long, steady increase of its concentration in the atmosphere was interrupted between 1999 and 2006. Bloom et al. (p. 322) use a combination of satellite data, which indicate water table depth and surface temperature, and atmospheric methane concentrations to determine the location and strength of methane emissions from wetlands, the largest natural global source. The constraints placed on these sources should help to improve predictions of how climate change will affect wet-land emissions of methane.

Abstract

Wetlands are the largest individual source of methane (CH4), but the magnitude and distribution of this source are poorly understood on continental scales. We isolated the wetland and rice paddy contributions to spaceborne CH4 measurements over 2003–2005 using satellite observations of gravity anomalies, a proxy for water-table depth Γ, and surface temperature analyses TS. We find that tropical and higher-latitude CH4 variations are largely described by Γ and TS variations, respectively. Our work suggests that tropical wetlands contribute 52 to 58% of global emissions, with the remainder coming from the extra-tropics, 2% of which is from Arctic latitudes. We estimate a 7% rise in wetland CH4 emissions over 2003–2007, due to warming of mid-latitude and Arctic wetland regions, which we find is consistent with recent changes in atmospheric CH4.

The atmospheric concentration of methane (CH4), an important greenhouse gas, is determined by a balance between natural and anthropogenic sources and sinks (1), leading to an atmospheric lifetime of approximately 9 years (2). Renewed interest in global budget calculations of CH4 levels is due to (i) the largely unexplained stability of CH4 concentrations during 1999–2006 and the renewed growth since early 2007 (3); (ii) laboratory and field measurements that support a small, previously unidentified, aerobic source of CH4 from terrestrial vegetation (4); and (iii) new satellite observations that provide additional constraints on current understanding (5). Concentration measurements of CH4 provide global constraints for emission estimates, but without additional, independent information it is difficult to attribute observed variability to individual sources and sinks.

Emissions from wetlands are the largest single source of CH4, representing 20 to 40% of the total CH4 emissions budget (1), of which 70% is estimated to originate from southern and tropical latitudes (6). Rice cultivation accounts for 6 to 20% of global CH4 emissions (1), the majority of which originates from south and southeast Asia (7). Methanogenesis, the biogenic production of CH4, occurs in natural wetlands and rice paddies by the anaerobic degradation of organic matter by methanogenic archaea. Production rates are controlled by the availability of suitable substrates; alternative electron acceptors for competing redox reactions, such as sulfate reduction (8); temperature; and soil salinity (9). Aerobic oxidation of CH4 by methanotrophs is a key factor in controlling CH4 emissions (10), with net fluxes to the atmosphere being primarily determined by the balance between CH4 production and consumption in the wetland soils. Emergent wetland vegetation can also increase the transport of CH4 between the soil and atmosphere (11). Although the controls on methanogenesis from wetlands and rice paddies are similar, the two sources are typically spatially distinct (12). Nevertheless, there is substantial uncertainty and regional variation associated with all these controlling factors. Wetland emissions dominated the interannual variability of CH4 sources over 1984–2003 (13). A decrease in wetland emissions over the past decade has reportedly masked a coincident increase in anthropogenic emissions (13), leading to stable global mean CH4 concentrations (14). Changes in the OH sink during 2006–2007 were not large enough to explain observed changes in CH4 concentration (3).

We present an approach to understanding the role of wetlands and rice cultivation in producing observed CH4 concentrations, using spaceborne measurements of gravity and CH4 over the 3-year period from 2003 to 2005. We used three data sets. First, we used satellite column observations of CH4 from the Scanning Imaging Absorption Spectrometer for Atmospheric Chartography (SCIAMACHY) instrument (15) aboard the Envisat satellite, which have been retrieved from solar-backscattered radiation at wavelengths from 1630 to 1679 nm (5), accounting for new water spectroscopic parameters (16). Retrieved columns, which are most sensitive to CH4 in the lower troposphere (5), range from 1630 to 1810 parts per billion, with the largest values generally over mid-latitude and tropical continents (16).

Second, we used gravity anomaly measurements from the Gravity Recovery and Climate Experiment (GRACE) satellite (17). These measurements, used in previous studies to investigate changes in groundwater, have been corrected for geophysical mass variations such as tides, atmospheric pressure, and wind (18). Relative equivalent water height Γ (in meters), inferred from gravity [see supporting online material (SOM)], shows seasonal variability ranging from 5 to 20 cm over major river basins (19). We used a Γ data set with a 10-day time step (18), which we regridded to 3° × 3°. Finally, we used surface skin temperature fields TS (in kelvin) from the National Centre for Environmental Prediction/National Centre for Atmospheric Research (NCEP/NCAR) weather analyses (20) as a proxy for soil temperature (SOM). We resolved all three data sets at the temporal and spatial resolutions of the Γ data set (SOM).

We find that that changes in wetlands and rice emissions dominate the observed variability of CH4 columns over wetland regions [square of the correlation coefficient (r2) = 0.7, SOM], and hence we interpret changes in these columns as changes in surface sources. We find that seasonal variations in the OH sink (21) and the CH4 source from fires (22) typically explain <10 and 3% of the observed CH4 column variability, respectively. CH4 column data are available only over cloud-free daytime scenes; changes in controls on wetland CH4 emissions on time scales shorter than 1 or 2 days due to processes such as rainfall, associated with cloudy conditions, are not well described by GRACE or Envisat. We excluded analysis over oceans, deserts, and regions with permanent ice cover.

To quantify the role of wetlands and rice cultivation in determining the observed variability of column CH4, we correlated these data with concurrent changes in Γ and TS over 2003–2005 (Fig. 1). We find that changes in Γ explain between 40 and 80% of the observed variability in CH4 measurements over the tropics. We find high correlations over many major river basins (SOM), with the exception of the Amazon basin, which is described below. We generally find a negative correlation between Γ and CH4 at high latitudes, which can be explained by high Γ in winter due to snow accumulation and associated low CH4 emission, and low Γ in spring and summer due to displacement of snow melt and higher CH4 emission as the exposed wetland is progressively warmed. At higher latitudes, we find that observed variations in CH4 are mostly explained by changes in TS (used here as a proxy for soil temperature). Changes in TS over the tropics explain little of the observed variation in CH4. Analysis of the deseasonalized time series shows similar but reduced correlations between CH4 and Γ and TS (SOM). This analysis provides global observations of the latitude dependence of the controlling factors—water table depth and soil temperature—that determine large-scale variations in wetland and rice paddy CH4 emissions (6). This work supports our model calculations (SOM) that show that wetland and rice paddy emissions are largely responsible for observed CH4 column variations.

Fig. 1

Correlations (r2) between cloud-free SCIAMACHY CH4 column volume mixing ratios (VMRs) (in parts per million) and (A) equivalent groundwater depth (in meters), determined from gravity anomaly measurements from the GRACE satellites (18) and (B) NCEP/NCAR surface skin temperatures (in kelvin), calculated on a 3° × 3° horizontal grid over 2003–2005. The correlation at a given point is determined by at least 15 and typically 60 CH4, groundwater, and temperature measurements. See SOM for a description of individual data sets.

Although variations in methanogenesis are predominantly attributed to variations in either groundwater or temperature, we account for the more complex dependence of methanogenesis with respect to both quantities (23). Within tropical latitudes, Γ is expected to be the dominant term in areas with distinct dry and wet seasons. In areas where the preexisting groundwater volume is large with respect to Γ variations, a combined Γ-TS relationship is expected. Figure 2 shows time series over four regions that exemplify the relationship between changes in Γ and column CH4. For the Niger and the Ganges basins, changes in Γ coincide closely with the CH4 variability, as is expected if the CH4 signal is due to methanogenesis. Over the Amazon basin, the overall correlation between Γ and CH4 is negligible (r2Amazon = 0.01). Changes in Γ over the Amazon basin are much larger than values observed over other river basins (Fig. 2) and lag behind CH4 changes by 1 to 3 months in the north of the basin, possibly due to the seasonal migration of the intertropical convergence zone (SOM), but we find a statistically significant correlation over the southern half of the basin (south of 4°S, r2 = 0.07). Although the CH4 seasonal cycles over the north and south Amazon are synchronous, the seasonal cycle of wetland groundwater over the north Amazon precedes the south Amazon cycle by approximately 2 months; considering the east-west divide of the Amazon basin does not improve the correlation. Wetland emissions over the Amazon basin coincide with the Amazon River system and its varzeas (24). We acknowledge that even large temporal changes in wetland groundwater, Γ(t), over this basin will not necessarily represent large changes in surface soil moisture because of the depth of the wetland groundwater, D + Γ(t), where D represents the initial volume of the water column.

Fig. 2

Time series of SCIAMACHY CH4 column VMR and groundwater depth over the (A) Ganges, (B) Niger, (C) South Amazon, and (D) South Congo river basins. The correlation (r2) between the variables is given for each panel. River basins are geographicaly defined with total runoff-integrating pathways (26). Vertical lines denote the start and end of each calendar year. A spatial representation of river basin correlations between CH4 and groundwater is included in the SOM.

To determine the distribution of wetland emissions of CH4 (Embedded Image, in mg/m2/day), we developed a simple model (SOM) that describes the time-dependent relation between these emissions and TS, and D + Γ(t) Embedded Image(1)where α is the fractional influence of Γ(t) on the total wetland groundwater volume D + Γ(t) (where 0 < α < 1); Q10(TS) describes the change in methanogenesis rate with a 10 K increase in temperature, where T0 is a constant (T0 = 273.16 K) (23); and k (mg/m3/day) incorporates other controlling factors (such as soil pH). The temperature dependence of Q10(TS) can be approximated by Q10(T0)[(T0)/(TS)] (23). We acknowledge that the derived values of Q10(TS) represent the relation between methanogenesis and TS as opposed to soil temperature (SOM). We maximized the local linear correlation between Embedded Image and SCIAMACHY CH4 columns by varying (D/α) on a per grid basis and globally fitting Q10(T0), where the gradient is proportional to changes in wetland emissions and the intercept is the sum of the remaining sources and sinks (SOM). We expect wetland and rice paddy emissions to follow a similar seasonal cycle, reflecting necessary hydrological and temperature conditions, but acknowledge that rice paddy emissions occur at specific intervals during the rice cultivation process. The global value of Q10(T0) that best fits the data is 1.65 ± 0.15, although we find that wetland and rice paddy emission distributions remain similar within the range 1 < Q10(T0) < 2.

The resulting normalized Embedded Image distribution was then scaled to a total global wetland and rice paddy source of 227 Tg of CH4/year, using the median value from the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (1) to derive global emission rates shown in Fig. 3A. We find the largest CH4 fluxes over South America, equatorial Africa, and southeast Asia. Emissions over the extratropical Northern Hemisphere are generally lower, but have elevated values over northern Europe and central Siberia and local peaks over North America. We find that uncertainties associated with extratropical CH4 fluxes are an order of magnitude smaller than those associated with tropical fluxes (SOM).

Fig. 3

(A) Logarithmic representation of wetland daily emissions of CH4 per unit of area inferred from fitting a temperature-groundwater wetland model to SCIAMACHY CH4 concentrations averaged on a 3° × 3° grid over 2003–2005. The normalized wetland and rice paddy emission distribution was scaled to 227 Tg of CH4 (1). (B) Zonal integral of bottom-up emission model estimates of CH4from wetlands, including bogs and swamps, and rice paddies (27) (red); from rice paddies only (green); and from normalized top-down CH4 emissions over 2003–2005 (blue). The shaded area indicates the uncertainty of our estimates due to systematic and random errors (SOM). (C) Predicted changes in annual wetland emissions for global wetlands, the tropics, the mid-latitudes from 23°N to 45°N, the mid-latitudes from 45°N to 67°N, the Arctic latitudes (>67° N), and the Southern Hemisphere. We assume a global wetland CH4 flux of 170 Tg/year in 2003 (1). The line thickness denotes the estimated uncertainty of the predicted changes, including random errors from Γ and TS measurements, and the error associated with 170 Tg/year, which we estimate as the standard deviation of global wetland CH4 emission estimates taken from the IPCC Fourth Assessment Report (1).

We used prior information about rice paddy distributions (12) to isolate wetland regions from our emission estimates. The resulting latitudinal distribution of wetland emissions is similar to those produced by independent bottom-up emission estimates (Fig. 3B) and is within the range of the large intermodel differences (25). We find that the tropics account for 55.5 ± 2.5% of global wetland emissions, with the Amazon and Congo river basins accounting for 20.0 ± 2.6 Tg of CH4/year and 25.7 ± 1.7 Tg of CH4/year, respectively. We find that rice paddy areas account for 29.1 ± 0.6% (66.0 ± 1.4 Tg of CH4/year) of the total rice plus wetland CH4 source, acknowledging that a small proportion of this may be attributed to the spatial coincidence of rice paddies and wetlands. We find that rice paddy emissions centered over China and south and southeast Asia account for 32.5 ± 3.7 Tg of CH4/year of the global rice paddy source, which is in agreement with bottom-up emission estimates (12).

We used our Embedded Image model to determine the evolution of wetland CH4 emissions over 2003–2007 relative to 2003 emissions. The change in annual emissions over that 5-year period was evaluated using the product of the fractional emission change and the wetland CH4 map in Fig. 3A. We omitted areas of rice cultivation (12), where year-to-year changes in CH4 emissions are determined by irrigation and other management regimes. We find a progressive global increase in CH4 from wetlands over 2003–2007, due mainly to temperature increases at extratropical latitudes (45° to 67°N). We also find that Arctic wetland emissions (>67°N) increased by 30.6 ± 0.9% over 2003–2007 to approximately 4.2 ± 1.0 Tg of CH4/year (SOM). We find that emissions from tropical wetlands remained constant over 2003–2006, with the exception of a 2.1 ± 0.7 Tg/year increase during 2007, most of which is accounted for by increased fluxes over the Congo (0.7 ± 0.2 Tg of CH4/year) and Sahel (0.9 ± 0.2 Tg of CH4/year) regions, as a result of increasing groundwater volume. The declining groundwater volume over tropical river basins over 2003–2006 did not significantly affect year-to-year changes in global wetland emissions. Our emissions calculations lead to better agreement with observed surface CH4 anomalies over 2003–2008 than those obtained using bottom-up wetland emissions (SOM), reproducing the observed post-2006 positive anomaly in both the Northern and Southern Hemispheres. This supports the idea that changes in wetland emissions have significantly contributed to recent changes in atmospheric CH4 concentrations.

There is substantial potential for wetland emissions to feed back positively to changes in climate (23), and therefore it is critical that we understand the extent of overlap between wetlands and regions that are most sensitive to projected future warming. We anticipate that the new constraints developed here will ultimately improve model predictions of this feedback.

Supporting Online Material

www.sciencemag.org/cgi/content/full/327/5963/322/DC1

SOM Text

Figs. S1 to S6

Table S1

References

References and Notes

  1. We thank J. Melack for providing feedback on the manuscript and R. Hipkin and F. Simons for assistance with GRACE gravity data. This work is funded by United Kingdom Natural Environmental Research Council studentship NE/F007973/1 and the National Centre for Earth Observation.
View Abstract

Stay Connected to Science

Navigate This Article