4.1. Spatial and Seasonal Variations in CH4 Emissions From the Newly Created Marshes
 Mean (SD) CH4 emission rate from the newly created marshes of the TGR was 6.7 ± 13.3 mg CH4 m−2 h−1 (ranging from −0.69 to 104.3 mg CH4 m−2 h−1) in the growing season of 2008. This was much higher than that from boreal littoral marshes, higher than that from rice paddies (ranging from 0.03 to12.7 mg CH4 m−2 h−1) in the same region [Han et al., 2005], and close to temperate marshes and tropical floodplains (Table 3). We did not measure CH4 emissions from the lake surface, though. Through comparing to surfaces of three tropical reservoirs (5.1 ± 5.8 mg CH4 m−2 h−1 for Petit Saut, 1.4 ± 2.0 mg CH4 m−2 h−1 for Balbina, and 3.3 ± 3.9 mg CH4 m−2 h−1 for Samuel) and a temperate lake (−2.5 to 5.7 mg CH4 m−2 h−1), which are all lower than that of the present study, therefore, we understood that the CH4 emission rate of the drawdown area should be higher than that of the reservoir surface of the TGR [Guérin et al., 2006; Wang et al., 2006]. On the basis of the average CH4 emission from Petit Saut, Balbina, and Samuel, we assumed the CH4 emission from the water surface of the TGR as 3.3 mg CH4 m−2 h−1. Calculating with the surface area (1080 km2), we further estimated the total CH4 emission from the surface as 3.6 Mg CH4 h−1. With the mean CH4 emission value of 6.7 ± 13.3 mg CH4 m−2 h−1 in the marsh area of TGR (100 km2 in area), the total CH4 emission from the marsh area of TGR could be estimated as about 0.67 Mg CH4 h−1, 19% of that from the surface. Considering the fact that the area of newly created marshes is only about 10% of the surface area, the newly created marshes in the drawdown area could be a “hotspot” of CH4 emission.
Table 3. CH4 Emissions From Littoral Marshes in Comparison With Similar Wetlands in Different Climates
|Finland, Lake Mekrijärvi and Heposelkä||Littoral marshes||−0.2–14.2||Juutinen et al. |
|Finland, Lake Ekojärvi||Littoral marshes||1.1–27.2||Kankaala et al. |
|Finland, Lake Alinen Rautjärvi||Littoral marshes||0–3.8||Kankaala et al. |
|China, Meiliang Bay in Taihu Lake||Littoral marshes||−1.7–131||Wang et al. |
|China, TGRR||Littoral marshes||−0.69–104.3||This study|
|Brazil, in Pantanal||Floodplains (flooded seasonally)||0.04–91.1||Marani and Alvala |
 Furthermore, Whiting and Chanton  showed that CH4 emissions from wetlands are controlled by primary production (PP), in other terms, CH4 emission is correlated to a net CO2 uptake. Evidences have also showed that approximately 3% of carbon fixed by photosynthesis at peak growing season was emitted as methane from wet sedge tundra [King and Reeburgh, 2002; King et al., 2002]. In our studying area, after impounding, the dominant ecosystems in the drawdown area transmitted from subtropical forests (known as a CH4 sink) to seasonally flooded freshwater marshes. Because of the limited exposure time (June to October) of the drawdown area, PP in the drawdown area might decrease after impounding. If we do not consider CH4 produced by plants under aerobic conditions [Keppler et al., 2006], though PP may be lower after impounding, because of the seasonal waterlogging and warm subtropical climate, the drawdown area may have been transitioned from a net CH4 sink to a net CH4 source.
 In the littoral marshes, CH4 emissions differ considerably within even a short (1–50 m) distance [Juutinen et al., 2001; Kaki et al., 2001]. Significant spatial variations of CH4 emissions were also found in the littoral marsh of the TGR (Figure 2). The high short-scale spatial heterogeneity suggests the need of fine-scale investigation. Like many other studies [Bubier et al., 1993; Ding et al., 2002], the present study showed a positive correlation between CH4 emission and standing water depths (Figure 4). Moreover, among the four plant stands of this study, the sequence of CH4 emission was accorded with the sequence of standing water depths (water table) (Table 1 and Figure 2). The correlation between CH4 emissions and vegetation characteristics (aerenchymal shoots; below-ground biomass; above-ground biomass) has also been found in many studies [Bubier et al., 1993; Greenup et al., 2000; Saarnio and Silvola, 1999; Van den Pol-Van Dasselaar et al., 1999]. In our study, plant biomass was not found to be significantly correlated with CH4 emission (p = 0.159). However, we found that CH4 emissions were significantly correlated with DOC content (Figure 3), which was partly derived from plants. The correlation observed here may be because of the fact that both pore water DOC and CH4 are produced by intense microbial activity in the sediments.
 In this study, we found a special seasonal variation of CH4 emissions, i.e., maximal emissions in early July, immediately after the exposure of the drawdown area, followed with a comparatively low and steady value in the rest of sampling period. Such seasonal pattern of CH4 emissions was different from those of natural marshes, including boreal peatlands, littoral zones of boreal lakes, alpine wetlands, in which the beginning of growing season shows a comparatively low emission, while the peak emission is recorded in the peak growing season (mid-July to August) [Alm et al., 1999; Chen et al., 2008; Kankaala et al., 2004]. The reason may be that the annual summer drought (from mid-July to mid-August in this area) limits CH4 emissions in July and August. In the drier period, CH4 emission rates would be at a low level because of low standing water depths.
 In our study, water temperatures were found to greatly influence the seasonal variations of CH4 emissions (Figure 4). The seasonal variations of CH4 emissions are ascribed to the seasonal balance of CH4 production, oxidation and transportation. The main factors which may influence the CH4 production rates over the growing season are changes in temperature [Yavitt et al., 1987], the amount of available organic substrate [Saarnio et al., 1997] and the size of the active anaerobic microbial biomass [Bergman et al., 2000]. Seasonal variations of CH4 oxidation and transportation have been reported to be correlated with temperature, light intensity, water level changes [Cicerone and Oremland, 1988] and seasonality in the functioning or physiology of the plants [van der Nat and Middelburg, 1998].
4.2. Implications for Large Dam Reservoirs
 In the recent ten years, scientists have gradually realized the role of large dams in emitting a substantial amount of GHGs, especially as “virtual CH4 factories” [Abril et al., 2005; Bambace et al., 2007; Fearnside, 2002; Fearnside, 2004; Giles, 2006; Guérin et al., 2006; Saint Louis et al., 2000]. A little disappointingly, the importance of dam-generated CH4 has largely been overlooked [Cullenward and Victor, 2006].
 Traditionally, dam CH4 budget is consisted of CH4 released at dam reservoir surface, turbines (of hydropower projects), spillways, and downstream [Abril et al., 2005]. However, created marshes in the drawdown area of the dam reservoir have been overlooked as a CH4 emission source of large dams [Abril et al., 2005; Fearnside, 2005; Guérin et al., 2006; Lima et al., 2008; Saint Louis et al., 2000]. In the present study, we preliminarily estimated that the CH4 emission of littoral marshes (covering only 10% of the surface area of TGR) was about 19% of the total CH4 emission from the surface of TGR. Therefore such newly created marshes should never be a negligible source of dam-generated CH4, considering 1) a higher CH4 emission rate comparing with that from the reservoir surface and 2) the great area of the drawdown zones of dam lakes which cover about 26,000 km2, estimated with an assumed low cover rate 10% of the drawdown zone of the dam reservoir and the total area of 0.26 million km2 for the global surface area of all reservoirs [Downing et al., 2006]. Moreover, a reasonable upscaling estimate for such ignored CH4 source could be attained on the basis of more data about CH4 emission from created marshes in the drawdown zones of reservoirs and detailed information about the drawdown area and drawdown duration of different dam reservoirs.