Glacial wetland distribution and methane emissions estimated from PMIP2 climate simulations

Abstract

The interglacial–glacial decrease in atmospheric methane concentration is often attributed to a strong decline in the wetland source. This seems consistent with the extreme coldness and vastly expanded ice sheets. Here we analyse coupled model simulations for the last glacial maximum from the Paleoclimate Modelling Intercomparison Project, using simple relations to estimate wetland characteristics from the simulated climate and vegetation. It is found that boreal wetlands shift southward in all simulations, which is instrumental in maintaining the boreal wetland source at a significant level. The mean emission temperature over boreal wetlands drops by only a few degrees, despite the strong overall cooling. The temperature effect on the glacial decline in the methane flux is therefore moderate, while reduced plant productivity contributes equally to the total reduction. Moisture effects play a role on the local scale only, while averaging out globally.

Keywords:

• climate
• modelling
• wetlands

1. Introduction

Past atmospheric methane concentrations have varied considerably, showing a decrease by more than 50% from the Pre-Industrial Holocene (PIH; 1850 AD) to the last glacial maximum (LGM; 21,000 years ago). Is this decrease due to major reductions in wetland area during the LGM and thus in wetland CH ₄ emissions? Recent bottom-up Earth System modeling studies have found only moderate reductions of 16–29% in LGM wetland emissions (Kaplan 2002; Valdes et al. 2005; Kaplan et al. 2006). In contrast, a recent top-down study postulates that the boreal wetland source was completely shut-down during the LGM (Fischer et al. 2008).

The present article addresses the question of whether large changes in climate during the LGM (surface cooling was 2–5°C in the tropics and 10–20° C in the Northern Hemisphere (NH) extratropics) can be concurrent with moderate changes in total wetland emissions. We do this by analysing climate model output from eight simulations that have been carried out in the second phase of the Paleoclimate Modelling Intercomparison Project (PMIP2; Braconnot et al. 2007). PMIP2 (http://pmip2.lsce.ipsl.fr/pmip2) has used fully coupled Atmosphere-Ocean and Atmosphere-Ocean-Vegetation General Circulation Models (AO and AOV-GCMs). Wetland area and emission strength are estimated here from the simulated LGM and PIH climate and vegetation, using relations derived from the literature. Wetland locations are determined by an algorithm based on soil moisture and temperature (Kaplan, 2002; Shindell et al. 2004). The wetland emission strength E (in mg/m²/day) is computed as follows:

Here k is a tunable constant, C is the fractional wetland coverage, NPP is net primary productivity, T is the soil temperature (in K) and Q ₁₀(T ₀) is a constant (Christensen et al. 2003; Gedney et al. 2004). The reference temperature is defined as T ₀ = 273K.

Estimates of the global annual methane flux from wetlands during the PIH vary widely. We chose a target value of 150 Tg, close to the bottom-up estimate of Houweling et al. (1999), and tuned the constant k in Equation (1) to give this global methane flux from wetlands. There is some discussion on which quantity best represents substrate availability for methanogenesis (Christensen et al., 2003). We choose NPP (see Walter et al. (2001); Kaplan (2002); Valdes et al. (2005)), rather than carbon content (Gedney et al., 2004), for the pragmatic reason that NPP is available from the PMIP2 database for two AOV-GCMs. The NPP from one model, HadCM-veg, was used in the emission computations for all models. Results are very similar when NPP from the other model, ECHAM-veg, are used.

2 Pre-industrial and glacial wetland methane emissions

The global methane flux from wetlands ranges 145–151 Tg in the different simulations (Table 1). The latitudinal distribution shows a minor peak between 40 and 60° N and a major peak around the equator (Figure 1, upper graph). Emissions from latitudes south of 30° S are negligible. The northern and tropical peaks in emissions reflect corresponding peaks in wetland area, but modified by the local emission strengths: boreal wetlands are less efficient than tropical wetlands, because they typically have lower temperatures and lower NPP. The latitudinal and seasonal distribution of emissions agree well with observations (e.g. Prigent et al. (2007)), taking into account that wetland area is assumed to have been ∼ 20% larger during the pre-industrial period than today (Chappellaz et al. 1993).

Figure 1. Annual methane emissions for the PIH (upper graph), the glacial changes (LGM minus PIH; second graph), emissions from PIH wetlands that were covered by ice during the LGM (third graph) and from LGM wetlands on continental shelves that were inundated during the PIH (lower graph) for eight different models. Emissions are integrated zonally and by 10° latitude belts.

Table 1. The PIH and LGM simulations included in the analysis, their type (atmosphere-ocean or atmosphere-ocean-vegetation), and the annual methane emissions from wetlands (in Tg) during the PIH and the LGM estimated from each model.

		PIH			LGM
Model	Type	Total	90–30N	30N–30S	Total	90–30N	30N–30S
HadCM3M2	AO	151	41	111	89	17	73
HadCM3M2-TRIFFID	AOV	145	41	105	85	15	70
CCSM3.0	AO	149	31	119	105	16	90
MIROC3.2.2	AO	150	37	116	87	15	74
ECHAM-MPIOM127	AO	146	21	127	93	8	86
ECHAM-MPIOM127-LPJ	AOV	147	21	128	91	10	82
IPSL-CM4-V1-MR	AO	150	47	104	97	17	81
CNRM-CM33	AO	148	31	119	92	12	81
Global emissions, NH extratropical and tropical values are shown.

Compared to the PIH, the LGM methane emissions show an overall decrease by 29–42% (Figure 1 and Table 1). Tropical emissions are reduced by about one-third (35–46 Tg) in the majority of models. Simulated reductions in boreal emissions are larger in a relative sense, ranging 51–65%, but smaller in absolute sense (11–30 Tg).

The reduction in methane emissions during the LGM can be due to climatic factors, temperature and soil moisture, changes in vegetation (NPP) and geographic effects such as where continental ice caps cover potential wetland area and new wetlands forming on the exposed continental shelves. Here we assume that the LGM is an equilibrium state and wetlands have formed on the exposed continental shelves. Ice sheets reduce boreal emissions, while continental shelves are found to dominantly affect tropical emissions (Figure 1, lower two graphs). Quantitative estimates of their separate impacts vary among models, but the combined effect on the global flux is consistently found to be small.

The climatic and vegetation effects can be separated from the geographic effects by considering a ‘common’ area where wetlands can potentially form during both time periods, thereby excluding the area covered by ice during the LGM and the land that is below sea level during the PIH. A factor analysis for this ‘common’ area shows that temperature and vegetation effects contribute equally to the total decline in the wetland source. Moisture changes can be important locally, but average out in the large-scale mean. For a full analysis, see Weber et al. (2010).

3 Redistribution of boreal wetlands

Temperatures decrease by 5–15°C in the region of boreal wetlands (here defined as the NH extratropics) and a large fraction of this region is coverd by the LGM ice sheets. Overall there is moderate drying, although some models show some areas which become wetter during the LGM. All models simulate moderately reduced boreal emissions (Table 1), nothwithstanding the pronounced cooling and loss of area. The reason for this is that the mean emission temperature is only a few degrees lower during the LGM as compared to the PIH (Figure 2), due to a southward shift in the location of wetlands. LGM emission temperatures reduce most over eastern Eurasia, while they are by and large similar to PIH temperatures over the west where there is a southward shift by as much as 15°. The small-scale LGM wetlands in northern America have varying emission temperatures compared to their more extended PIH counterparts. This redistribution of boreal wetlands is due to a southward displacement of the westerlies by the Laurentide and Fennoscandian ice sheets and a southward extension of snow and soil freeze in winter.

Figure 2. The mean temperature during the emission season, latitudinally averaged over latitudes 30–90° N, for the PIH (solid lines) and the LGM (dashed lines) in four PMIP2 simulations: CCSM3.0 (red: offset by −5°C), HadCMsM2 (black), HadCM3M2-TRIFFID (blue) and MIROC3.2.2 (green, offset by + 8°C).

In the tropics there is also a southward shift, related to a southward shift of the inter-tropical convergence zone. This mainly results in a redistribution of tropical wetlands, not in a change in the net soure.

4 To conclude

The present PMIP2-based estimate of the glacial wetland source is 29–42% of the pre-industrial value. This is a stronger reduction than found in earlier bottom-up modelling studies (Kaplan 2002; Valdes et al. 2005; Kaplan et al. 2006). The present PMIP2 climate model simulations show less continental drying and stronger cooling in the tropics that those used by earlier studies. This results in a pronounced reduction in the glacial tropical source. The present results do not support a complete shut-down of the boreal source during the LGM. This is due to the relocation of boreal wetlands enabled by the moderate continental drying.

The relative reduction in the glacial atmospheric CH₄ concentration is expected to be somewhat larger than the relative reduction in the source, because of chemical feedbacks and other factors affecting CH ₄ lifetime. Model studies find lifetime to be reduced by 10–20% during the LGM (e.g. Kaplan et al. (2006)). Variations in other methane sources than wetlands may have played a role too: anthropogenic sources are likely to have been significant well before 1850 AD, given the size of the world population at that time (discussion in Chappellaz et al. (1997)). Taking this into account, the present estimate of the glacial wetland source is easily reconciled with the observed drop in CH₄ concentration. The presently found distribution in emissions is also consistent with the low interpolar difference in glacial methane concentrations derived from ice-core data.

Acknowledgements

The authors acknowledge the international modeling groups for providing their data for analysis and the Laboratoire des Sciences du Climat et de l'Environnement (LSCE) for collecting and archiving the model data. The PMIP2 Data Archive is supported by CEA, CNRS, the EU project MOTIF (EVK2-CT–2002–00153) and the Programme National d'Etude de la Dynamique du Climat (PNEDC).