Progress in the studies on the greenhouse gas emissions from reservoirs

The green credentials of hydroelectricity in terms of greenhouse-gas (GHG) emissions have been tar-nished with the ﬁnding of the researches on GHG emissions from hydroelectric reservoirs in the last two decades. Substantial amounts of GHGs release from the tropical reservoirs, especially methane (CH 4 ) from Brazil’s Amazonian areas. CH 4 contributes strongly to climate change because it has a global warming potential (GWP) 24 times higher than carbon dioxide (CO 2 ) on a per molecule basis over a 100-year time horizon. GHGs may emit from reservoirs through four different pathways to the atmosphere: (1) diffusive ﬂux at the reservoir surface, (2) gas bubble ﬂux in the shallow zones of a reservoir, (3) water degassing ﬂux at the outlet of the powerhouse downstream of turbines and spillways, and (4) ﬂux across the air–water interface in the rivers downstream of the dams. This paper reviewed the productions and emissions of CH 4 , CO 2 , and N 2 O in reservoirs, and the environmental variables inﬂuencing CH 4 and CO 2 emissions were also summarized. Moreover, the paper combined with the progress of GHG emissions from Three Gorges Reservoir and proposed three crucial problems to be resolved on GHG emissions from reservoirs at present, which would be beneﬁt to estimate the total GHG emissions from Three Gorges Res-ervoir accurately. (cid:2) 2014 Ecological Society


Introduction
Carbon dioxide (CO 2 ), methane (CH 4 ), and nitrous oxide (N 2 O) are the three principal greenhouse gases (GHGs) in the atmosphere, and continuously increases in atmospheric concentrations of three GHGs are closely related to global climate change [1]. The studies on the GHG emissions from reservoirs in the last two decades indicated that hydroelectricity was not a green and clean energy as expected that no GHG is emitted from the reservoir surface [2][3][4]. In fact, reservoirs are also an important GHG source in the terrestrial ecosystems [5,6]. According to the natural belts that reservoirs located, the global reservoirs could be divided into tropical reservoirs (e.g., reservoirs in Brazil, French Guiana, and Laos) and temperate reservoirs (e.g., reservoirs in Canada, Switzerland, and China). The global warming potential (GWP) of the GHG emissions from Brazil's reservoirs are amazing, which are even higher than that from thermal power plants with similar installed capacity [2]. For example, Curuá-Una Reservoir in Brazil emitted 3.6 times more GHGs than those would have been emitted by generating the same amount of electricity from oil [7]. However, GHG emissions from Canadian reservoirs are relatively low [8], which are lower than the GHG emissions compared with GHGs emitted by fossil-fuelled electricity generation. Therefore, it cannot be generalized to determine whether the development of hydroelectricity could reduce GHG emissions, which should depend on the specific situation of reservoirs. The geographic locations of reservoirs have an impact on the organic matter storage and water temperature, and influence on CO 2 and CH 4 emissions subsequently [6]. However, CH 4 emission fluxes from Lake Wohlen, a temperate reservoir in Switzerland, are even higher than those from tropical reservoirs [9], which cause the controversy on the development of hydroelectricity in the middle Europe region [3]. Beside latitudes, CO 2 emissions from reservoirs are also influenced by reservoir ages [6], wind speeds [10], pH values [11], precipitation [12], chlorophyll-a concentrations [12,13], and dissolved organic carbon in the water body [12,14], while CH 4 emissions from reservoirs are influenced by water depths [15], water level fluctuations [16], DO concentrations [17], water velocities [16], and wind speeds [10].
GHG emissions from reservoirs are different from the natural water bodies, such as lakes and rivers, because the impoundment of the reservoir has resulted in flooding of large areas of terrestrial and natural aquatic ecosystems. CO 2 and CH 4 are the major end products of the microbial decomposition of flooded organic matter [17], which are transported to the atmosphere from the reservoir surface by diffusion or bubbles. Turbines and spillways are unique to the dams, and turbines are used to generate electricity by transforming potential energy of the storage water into electric energy by the rotation of vane wheel; spillways are the drainage channels to control the floods in the reservoirs. When the deep water passes through the turbines and spillways, the dissolved gas (especially CH 4 ) in the hypolimnion before the dams would release into the atmosphere, becoming a huge CH 4 source, because of the abrupt change in temperature and pressure, which is called ''degassing'' [18]. Besides, downstream fluxes are often higher than upstream ones because of the strong disturbance to the water passing through the dams [19]; thus, the downstream emission fluxes should be paid attention. In conclusion, there are 4 pathways for GHG emissions from reservoirs, i.e., diffusive emission, ebullitive emission, degassing emission at turbines and spillways, and downstream emission [20].
The CO 2 emission from reservoirs is the largest, the second is CH 4 emission, and N 2 O emission is the smallest. However, the GWP of the three gases is different. CH 4 has a GWP 24 times higher than carbon dioxide (CO 2 ) on a per molecule basis over a 100-year time horizon [3], and nitrous oxide (N 2 O) has a GWP 298 times that of CO 2 [21]. Based on the studies on GHG emissions from reservoirs available, this paper reviewed the 3 GHG emissions from the tropical and temperate reservoirs through diffusion, ebullition, degassing, and downstream river. In addition, the environmental variables influencing GHG emissions were also summarized.

CO 2 production in reservoirs
In a broad sense, CO 2 production in a reservoir includes the carbon footprint of emissions from the use of fossil fuel, steel, and cement during the construction phase of a dam [21], which is related to the size of dam and the duration of creation. The Three Gorges Dam (TGD) is a good example, with a length of 3035 m and a height of 185 m, which lasted for 18 years to construct (1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009). Although there is no study on CO 2 emission during the construction phase of the TGD, CO 2 emission during the process cannot be ignored. Besides, CO 2 production in a reservoir also includes the CO 2 emission when the dam operated normally, e.g., CO 2 emission from the fossil fuel combustion by shipping, and CO 2 emission from the turbines. Navigation and electricity generation are two important functions of the Three Gorges Reservoir (TGR), but CO 2 emission has not been quantified during the two processes by far. CO 2 discussed in the paper is produced from the decomposition of the flooded organic matter under the aerobic or anaerobic conditions after the impoundment. Carbon sources in the reservoirs included the flooded organic matter in the original forests, soils, vegetations, allochthonous input from terrestrial ecosystems or the upstream rivers nearby, and photosynthetic fixation by phytoplankton at the reservoir's surface or vegetations in the drawdown areas [21][22][23]. The flooded organic matter would decompose into CO 2 and CH 4 by methanogens under the anaerobic conditions at the reservoir bottom [23,24]. In fact, CO 2 could also be produced at the aerobic conditions, e.g., the decomposition of dead trees left above the water surfaces [24].

CO 2 transport in reservoirs
CO 2 emission fluxes in the reservoirs mainly include the two ways, i.e., diffusion and ebullition [24]. Diffusion is the dominate way for CO 2 emission from reservoirs [25], while bubbles have little contribution to CO 2 emission from reservoir's surface, because the solubility of CO 2 is large, i.e., 1 L water could dissolve 1 L CO 2 at the conditions of 1 atm and 25°C; thus, CO 2 is easily absorbed by water during the transport from the reservoir's bottom. For example, bubbles contributed less than 1% of CO 2 emission from diffusion during the first years after the impoundment for Petit Saut Reservoir, French Guiana [23]; the CO 2 diffusive emission from Brazil's Balbina Reservoir is 2450 Gg C a À1 , while the CO 2 ebullitive emission is only about 0.02 Gg C a À1 [26]. According to Table 1, bubbles are not the important way to transport CO 2 in tropical reservoirs, and only the CO 2 diffusive fluxes are studied in temperate reservoirs (Table 2), probably because the frequency of bubbles and CO 2 concentrations in bubbles are low and even could be ignored in temperate reservoirs.

Influences of turbines and spillways on CO 2 emission
The intakes of turbines and spillways often locate in the dozens of meters depth below the water surface, where have remarkable higher pressure than the atmospheric pressure. The dissolved CO 2 in the hypolimnion would be released into the atmosphere when the water passes through the turbines and spillways because of the abrupt decrease in pressure and the increase in water temperature [18]. Compared with the CO 2 diffusive emission from the reservoir surface, the CO 2 degassing fluxes at the turbines and spillways have relatively small contribution to the total CO 2 emission of a reservoir. Based on the literatures available [23,26,31,33], CO 2 degassing at the turbines and spillways contribute 0-16% to the total emission. Specifically, CO 2 degassing emissions from the turbines and spillways of Petit Saut Dam account for 7-16% of the total CO 2 emissions during 1994 to 2003 [23], which is called ''degassing proportion'' hereafter. The degassing proportion for CO 2 emission is about 1.67% at the Balbina Dam [26], 0.4-5% at the La Grande-2 Dam, Canada, and 0.1-7% at the La Grande-2 Dam [33], but the CO 2 degassing proportion is close to 0 at the Nam Ngum Dam and Nam Leuk Dam, Laos (Table 1) [31]. Up to now, no measurement has conducted on the degassing emission in China because of the limitations in the measurement technique and the military supervision for Chinese dams. However, the part of degassing emission is crucial to the total GHG emissions from reservoirs, especially for CH 4 degassing fluxes.

CO 2. emission from the downstream rivers
Based on all the studies available [19,23,26], CO 2 emission from the downstream rivers contributed to 1.63-32% of the total CO 2 emissions from the reservoirs. Specifically, CO 2 emissions from the Uatumã River, the downstream of the Balbina Dam, account for 1.63-7% of the total CO 2 emissions from the Balbina hydroelectric system [19,26]; CO 2 emissions from the Sinnamary River, the downstream of the Petit Saut Dam, contribute to 22-31% of the total CO 2 emissions from the Petit Saut hydroelectric system [19,23]; CO 2 emissions from the Jamari River, the downstream of the Samuel Dam, account for 32% of the total CO 2 emissions from the Samuel hydroelectric system [19]. CO 2 and CH 4 emissions at Sandouping, a downstream site of the TGD, are significantly higher than these emissions at Zigui, a site just upstream the TGD. The difference in the CO 2 and CH 4 emissions between the two sides of the TGD are probably related with the strong disturbance by the water passing through the turbines and spillways and the faster water velocity in the downstream rivers [34].

Environmental factors and their effects on CO 2 emission from reservoirs
2.5.1. Organic matter and temperature CO 2 and CH 4 emissions from the reservoir's surface are related with the amount of easily decomposable organic matter that is flooded after the filling of the reservoirs [35]. A large amount of organic carbon is stored in peatlands and forests; thus CO 2 and CH 4 emission fluxes are very high when such two land use types are flooded, such as Eastmain-1 Reservoir [36], La Grande 2 Reservoir [37], Laforge-1 Reservoir in Canada [37], and Brazil's Balbina Reservoir [26]. CO 2 and CH 4 emissions from reservoirs are very low if barren soils are flooded in the canyons, such as Ertan Reservoir [15], TGR [34]. Therefore, intensity clearing activities are often carried out before the impoundment of hydroelectric reservoirs in China, which could prevent the water quality after the water storage, and reduce GHG emissions as well. Fluctuation in the water temperature has an impact on the CO 2 solubility [22], primary production [12], and the decomposition in organic carbon. [12]. Elevation in the water temperature promotes CO 2 emissions by increasing the decomposition rate of organic carbon, which could be seen in the positive correlations between CO 2 emissions and water temperatures in Canadian reservoirs [38,39]. However, if there are algae distributed in the water surface, the elevation in the water temperature would promote CO 2 absorption because of the increase the primary production of aquatic plants [12].

Latitude and reservoir age
CO 2 emission fluxes are exponentially negative correlated with the latitudes of the geographic location of hydroelectric reservoirs [6]. Most of CO 2 emission fluxes in the tropical reservoirs are higher than those in the temperate reservoirs (Tables 1 and 2), because a large amount of organic carbon is stored in the tropical flooded areas, and the high water temperature is beneficial for the decomposition of organic matter [35]. CO 2 emission fluxes decrease with the increase of reservoir ages because of the gradual decrease in the storage of organic carbon in the reservoirs [35]. GHG emissions are mainly released at the initial periods after the impoundment due to the abrupt release of the nutrient substances in the flooded lands, the elevation of microbe activities, and the decomposition of unstable carbon matters, such as soils, litters, swigs, and leaves [39]. However, CO 2 emission fluxes gradually decrease in the reservoirs with the decrease of organic matter and the increase of CO 2 absorption through photosynthesis by various aquatic plants. Specifically, CO 2 emission fluxes reached up to 8000 mg CO 2 m À2 d À1 at the surface of Eastmain-1 Reservoir in 2006, the first year after impoundment [40], while decreased to 2426 mg CO 2 m À2 d À1 in 2009 (Table 2) [8]; the average CO 2 emissions from Petit Saut Reservoir were about 200 Â 10 3 t a À1 in the first 3 years after impoundment (1994)(1995)(1996), but decreased until less than 70 Â 10 3 t a À1 after 2000 [23].

pH value
CO 2 emissions or CO 2 partial pressure are significantly negative related with pH values in the reservoir's surface, as were seen in the five reservoirs in Wujiang River in China [48], the reservoirs in the western United States [11], and the reservoirs in eastern Canada [38]. The pH critical values are often reported to be 7.9-8.5 between CO 2 absorption and emission [11,38,39]. PH level influences the CO 2 concentrations in the water by favoring the formation of bicarbonate at the alkaline conditions, which leads to an undersaturation of dissolved CO 2 , promoting the absorption of atmospheric CO 2 [38].

Vegetation
Whether there are vegetations distributed in the reservoir's surface determines that the reservoir is CO 2 source or CO 2 sink. When aquatic plants are distributed in the reservoir's surface, such as alga, CO 2 would be absorbed from the air and water nearby by photosynthesis. For example, CO 2 sink during the low water level periods is related with algal blooms in Xiangxi River and Pengxi River in summer [13,45]. If there is no vegetation distributed in the water surface, it would be possibly CO 2 source to the atmosphere, because the CO 2 partial pressure in the water is higher than that in the atmosphere.

Wind speed
The gas transfer velocity is influenced by wind speeds at the air-water interface [49]. It is considered that the strong wind (larger than 3 m s À1 ) would promote the dissolved gases to release from the water surface, and the thin boundary layer model is appropriate to measure the gas fluxes under the conditions of strong winds [50,51]. When wind speed is less than 2-3 m s À1 , the gas transfer coefficient (k), an important parameter in the boundary layer model, is independent on wind speed [50], thus the error would be large if the empirical model is used to calculate CO 2 flux. Floating chambers are good estimates of CO 2 diffusive fluxes under the conditions of low wind speeds ranging from 0 to 3 m s À1 , because gas fluxes are not easily influenced by the walls of chambers at such conditions [11].
2.6. Summary CO 2 diffusive emissions are predominant in the upstream before a dam. CO 2 degassing emission from the turbines and spillways accounts for 0-16% of the total CO 2 emission from a hydroelectric system, and the downstream emission accounts for 1.63-32% of the total. To estimate the total CO 2 emission from a reservoir accurately, it needs to measure upstream CO 2 emission, degassing emission from turbines and spillways, and the downstream river. In addition, CO 2 emissions from reservoirs are influenced by the flooded organic carbon, water temperature, geographic location of reservoirs, reservoir age, pH value, vegetation, and wind speed.

CH 4 production in reservoirs
Damming for hydroelectric production involves the flooding of vegetations and soils containing significant amounts of organic matter. CH 4 , N 2 , and a small amount of CO 2 are released from the decomposition of various complicated organic matter under anaerobic conditions, while only CO 2 and N 2 are released under aerobic conditions [28]. Organic carbon is decomposed into CH 4 undergoing microbial fermentation, which could be divided into two steps as follows. Firstly, carbohydrate, fatty acid, and protein in the sediments of the flooded lands are decomposed into the simple organic acid (i.e., farmate, acetate), small molecular alcohols (i.e., methanol, ethanol, isopropanol), methylamine, and dimethyl sulfide by hydrolyzation and fermentation, and these simple organic matter could be decomposed further by methanogenesis [52]. Secondly, CH 4 and CO 2 would be released from the decomposition of these substances by methanogens under the anaerobic conditions, and the following reaction equation reflects the decomposition process from acetic acid to CH 4 :CH 3 COOH ? CH 4 + CO 2 . The two processes happened synchronously [25,53].

CH 4 transport in reservoirs
Ebullition is a dominant way for CH 4 emission, while molecular diffusion is a secondary way for CH 4 emission from tropical reservoirs (Table 1). Bubble emissions are independent on reservoir ages. For examples, CH 4 emission by bubbles is about 1-3 times as high as that by diffusion in Petit Saut Reservoir in the first 6 years (1994-1999) after filling [23]; CH 4 emission by ebullition is 3 times higher than that by diffusion in Curuá-Una Reservoir (20-21 years old) in the wet season [7]; 98% of CH 4 is released by bubbles in Gatun Lake, Panama, which is an old reservoir with 84 years old [27]. However, molecular diffusion is a dominant way for CH 4 emission from temperate reservoirs ( Table 2), but sometimes bubbles also have remarkable contribution to the total CH 4 emissions from some temperate reservoirs, and Lake Wholen in Switzerland, an old reservoir with ages of 90 years, is a representative example. CH 4 ebullitive fluxes are up to 480 mg CH 4 m À2 d À1 , and CH 4 diffusive fluxes are only about 15 mg CH 4 m À2 d À1 from Lake Wholen, which is by far the highest CH 4 emission rate ever documented for a midlatitude reservoir, even higher than that in tropical reservoirs (Tables 1 and 2) [9]. Inundated forests provide a large supply of nutrients for microbes in the Lake Wholen. Thus, the water in Lake Wohlen sometimes looks like champagne in the summer, with masses of gas bubbles rising to the surface [54]. In addition, recently studied indicated that the high sedimentation rate supplies reactive organic matter to deep, anoxic sediment strata at a very high burial rate (1100 g C m À2 a À1 ), which would fuel methanogenesis and ebullition of CH 4 at a warm water temperature (17°C) [55].

Influences of turbines and spillways on CH 4 emission
The DO concentrations decrease with the water depths, while the dissolved CH 4 concentrations increase with the water depths in a reservoir [17,19,23,56]. The dissolved CH 4 concentrations reach 7.5 mg L À1 at a depth of 30 m in Brazil's Tucuruí Reservoir [18]. When the water in the hypolimnion emerges from the turbines and spillways, the pressure instantly drops to a level of 1 atm, and the water temperature increases near to the temperature in the epilimnion; thus the great majority of the dissolved gas is released into the atmosphere of the downstream river because the solubility of CH 4 is only 0.035 mg L À1 at the normal pressure and temperature (1 atm, 25°C) [18]. CH 4 degassing emissions from the turbines and spillways are the dominant part of the total CH 4 emissions from a hydroelectric system. For examples, 64.9% and 34.5% of the total CH 4 emissions release from the turbines and spillways of the Tucuruí Dam, respectively, while CH 4 emissions from the reservoir's surface, including bubbles and diffusion, only account for 0.6% of the total CH 4 emissions [29]; CH 4 degassing emissions from the turbines contribute to 42.4-46.6% of the total CH 4 emissions from the Balbina Reservoir [30,57]; CH 4 degassing emissions from the turbines and spillways exceed 50% of the total CH 4 emissions from the Petit Saut Reservoir during the first 10 years after impoundment except for 1994, because CH 4 ebullitive emissions account for a large proportion of the total CH 4 emission in 1994, the first year after filling of the Petit Saut Reservoir [23]. In Brazil's other reservoirs, CH 4 degassing emissions from the turbines and spillways exceed the half of the total CH 4 emissions. For examples, 95% of the total CH 4 emission occurs at the turbines and spillways of the Curuá-Una Dam [7], 99% of the total CH 4 emission is degassing at the turbines and spillways of the Itaipu Dam [29], and 80.8% of the total CH 4 emission originates from the degassing emission from the turbines and spillways of at the Serra da Mesa Dam [29]. The turbine intakes of TGD are located at 80 m depth, and the pressure is about 8 atm there, but there is no report on the degassing emission from the turbines and spillways of the TGD by far.

CH 4. emission from the downstream river
Downstream CH 4 emission cannot be ignored in a hydroelectric system, because CH 4 diffusive flux is higher in the downstream river than the upstream. For example, the average CH 4 emission flux is (60 ± 38) mmol m À2 d À1 at the downstream rivers of Petit Saut, Balbina, and Samuel dams, which is significantly higher than the average upstream CH 4 emission flux ((3 ± 2) mmol m À2 d À1 ). However, rivers downstream of dams account for 9-33% of the total CH 4 emissions across the reservoir surfaces because of the relatively small areas of downstream rivers [19].

Environmental variables and their effects on CH 4 emissions from reservoirs
3.5.1. Temperature CH 4 is the end product of the anaerobic decomposition of organic matter by multiple microbes, and CH 4 emissions from the reservoirs are influenced by temperatures [38]. Elevation of temperature can increase the activities of microbes. Methanogenic bacteria are much more responsive to temperature than methanotrophic bacteria [52], and the optimum temperatures for CH 4 production and oxidation is about 25°C [58]. Therefore, when the temperature fluctuated at the range that two kinds of bacteria could bear, more CH 4 would be produced when temperature rises, and CH 4 production rate would decrease remarkable when temperature falls [58]. Therefore, CH 4 emissions are often reported to be linear or exponential relationships with soil temperatures or water temperatures [9,[59][60][61].

Water depth
Most of CH 4 would be oxidized by methanotrophic bacteria when the water passes through the O 2 rich epilimnion waters by bubbles or molecular diffusion from reservoir sediments, and only a small amount of CH 4 could release into the atmosphere ultimately. CH 4 is more easily emitted from the shallow-water areas compared with the deep-water areas, because less CH 4 would be oxidized through a short distance between the surface and the reservoir bottom [60]. Although the occurrence of bubbles is episodic [6], bubbles often occur in the areas with a depth less than 10 m [17,23,27,37]. The shallower the water is, the more CH 4 ebullitive flux emits. For examples, CH 4 ebullitive fluxes are only 10-200 mg CH 4 m À2 d À1 in the areas deeper than 7 m in Gatun Lake, Panama, while the values reach 300-2000 mg CH 4 m À2 d À1 in the areas less than 2 m [27]; the average CH 4 ebullitive flux was (164 ± 50) mg CH 4 m À2 d À1 in the areas with depths of 0-3 m, while the CH 4 bubble flux was 0 in the areas deeper than 8 m in Petit Saut Reservoir in 1997 [56]. Moreover, the CH 4 diffusive fluxes are also influenced by water depths. For examples, the CH 4 diffusive emissions from the water surface at the 1 m and 5 m depths are higher than those at the 10 m and 50 m depths in Ertan Reservoir [15]; the CH 4 diffusive fluxes in the areas less than 3.5 m are significantly higher than those in the areas deeper than 7 m in the La Grande-2 Reservoir and Laforge-1 Reservoir, Canada [37].

Water level fluctuation
Water level fluctuation in a reservoir changes the water depths of the permanently flooded lands, which influences on CH 4 emissions from the reservoir's surface subsequently. The drawdown areas are caused by the fluctuation in water levels in the edge of a reservoir [62], and the vegetations in the drawdown area are served as ''methane factory'' [21]. The vegetations grow and absorb CO 2 from the atmosphere by photosynthesis when the drawdown areas are exposed each year, but the new fixed carbon would be decomposed into CH 4 under the anaerobic conditions in the bottom of reservoirs when the drawdown areas are inundated again [63]. The vegetations in the drawdown areas can continuously remove carbon from the atmosphere as CO 2 and return it as CH 4 , with a much greater impact on global warming [63]. Such processes are repeated every year, but there is still lack of quantitative study in aboard. However, there are several studies available on CH 4 emission from the TGR drawdown areas [16,64,65]. After impoundment of the TGR, there is large drawdown areas distributed in the two sides of the TGR, with 30 m in height and 450 km 2 in area [34], which is seldom seen in the world. The difference is obvious in CH 4 emissions between drylands and wetlands during the drainage of the TGR drawdown areas. CH 4 emission fluxes in the natural wetland (e.g., Scirpus triqueter) and man-made wetland (e.g., rice paddies) reach up to 14.9 mg CH 4 m À2 h À1 [64] and 3.94 mg CH 4 m À2h À1 [16], respectively, while CH 4 emissions from the different types of drylands (i.e., fallow land, deforested land, cropland) are close to 0 during the drainage periods [16]. However, CH 4 emission fluxes are low both in the drylands and wetlands during the inundated periods (0.1-0.3 mg CH 4 m À2 h À1 ) [16].
3.5.4. Dissolved oxygen in the water CH 4 emission fluxes at the air-water interface are positively related with the dissolved CH 4 concentrations in the surface layer [56], while the dissolved CH 4 concentrations are influenced by DO in the water. Thus, the DO concentrations in the surface layer have an effect on CH 4 emission from the water surface. The DO concentrations decrease with the water depths, which decreases the possibility of CH 4 oxidation in the hypolimnion. Besides, another reason why the CH 4 concentration is high in the hypolimnion is that the deep waters are closed to the sediments, the CH 4 source of production [66,67].

Other environmental variables
CH 4 emissions are also influenced by wind speed [49,68], water velocity [69,70], and air-water temperature difference [71,72], which causes the instability of the air-water interface. Such 3 variables have an effect on CH 4 emission by changing the gas transfer coefficient (k) at the air-water interface [10,49,69]. Besides, CH 4 emissions from the reservoirs are also influenced by the flooded organic carbon [35], the carbon input from the upstream rivers [23,73], the primary production of aquatic plants [23], the retention time of water [73], weather conditions (light, rainfall) [74], and water quality [43,74].

Summary
Ebullitive emissions are the dominant way for CH 4 emission from the surface of tropical reservoirs, while diffusive emissions are the main way for CH 4 emission at the air-water interface in the temperature reservoirs. Moreover, CH 4 degassing emissions from the turbines and spillways account for the major proportion of the total CH 4 emission from a reservoir. Furthermore, CH 4 fluxes are high in the downstream river, which cannot be ignored when estimate the total CH 4 emission from a reservoir. In addition, CH 4 emissions from reservoirs are influenced by temperature, water depth, water level, DO, organic carbon, and so on.

N 2 O production in reservoirs
N 2 O could be released both in the aerobic or anaerobic conditions, but the substrates and bacteria are different between the two conditions. NO 3 À could reduce into NO 2 and N 2 via microbial denitrification under anaerobic conditions [76], while NH 4 + could oxidize into N 2 O and N 2 by nitrifying bacteria under aerobic conditions [77]. N 2 O, as the intermediate byproduct of two microbiological processes, could be produced in oxic epilimnion, in anoxic hypolimnion, at oxic-anoxic boundaries, either at the sedimentwater interface [78].

N 2 O emission flux in reservoirs
Diffusion flux is the main way for N 2 O emission from the reservoir surface, while bubbles have little contribution to N 2 O emission because of the high solubility of N 2 O. For example, only about 0-0.7 lg m À2 d À1 of N 2 O is emitted by bubbles in Lokka Reservoir of Finland, which could be ignored [43]. Moreover, upstream N 2 O emission flux at the reservoir surface is higher than that in the downstream rivers. For example, normalized N 2 O fluxes are 3-4 times higher at the upstream of Petit Saut Dam than those in its downstream river [79].
The impoundment of reservoirs may have little effect of adding additional N 2 O emission, while may even decrease N 2 O emission, because N 2 O emission from lakes is significantly lower than that from croplands [78]. The original soils before filling, especially croplands, are an important N 2 O source [76], but N 2 O emission fluxes decrease even become a weak N 2 O sink when the soils are flooded after impoundment [77]. However, it is not uniform for N 2 O sink or source at various land use types before impoundment. Therefore, only if N 2 O emission fluxes were measured at different land uses before and after impoundment in details, the net N 2 O emission could be calculated accurately.

Influences of turbines and spillways on N 2 O emission
The distribution of dissolved N 2 O concentrations at the vertical profiles is different from that of dissolved CH 4 concentration. The dissolved N 2 O concentrations increase with the decrease in DO, but the dissolved N 2 O concentrations are undersaturated in completely anaerobic layers, which is consumed due to microbial denitrification in the reservoir sediments [78]. However, there is no report on N 2 O degassing emission when the water passes through turbines and spillways.

Summary
There are a few studies available on N 2 O emissions from temperature and tropical reservoirs (Table 3)

Outlook
Several studies on GHG emissions from reservoirs are undertaken by Chinese scholars. Among these studies, GHG emissions from the TGR have drawn many attentions, and experts and scholars were organized to discuss several times in the past 5 years. The TGR is a typical valley-type reservoir, and the surface area of the TGR is up to 1084 km 2 , with 660 km in length and 1-2 km in width when the water level reached 175 m. Spatial variations in GHG emissions from the TGR are significant due to the heterogeneity in hydrological conditions [81][82][83]; thus, to avoid mistaking GHG emissions from the TGR, different research teams should strengthen the cooperation to reveal the patterns of GHG emissions from the TGR correctly.
GHG emissions from the TGR have been studied for over 5 years since the CH 4 emissions from the TGR drawdown areas aroused attention at home and abroad in 2009 [64,84]. Much progress has been achieved on CH 4 emissions from the open water areas and the drawdown areas in the TGR by far [16,34,64,65]. However, there are 3 main defects on the present research. Firstly, no study is conducted on the degassing fluxes at the turbines and spillways of the TGD, especially substantial amounts of CH 4 emission when the hypolimnion passes through the turbines and spillways. Secondly, the dissolved gases in the water could reflect the potential abilities of gases storage and emission from a reservoir, but there is no study on the dissolved concentrations of CO 2 , CH 4 , and N 2 O in the water of the TGR because of the limitation in technique. Thirdly, there are many tributaries distributed in the TGR region, and every tributary has different conditions in hydrology and water quality, but the studies on GHG emissions from tributaries are limited in Xiangxi River, Pengxi River, Longxi River, and Daning River [13,45,83]. Due to these defects, it is difficult to estimate the total GHG emissions from the TGR accurately based on the present data available.
Besides the TGR, China has many other large dams with height higher than 15 m. According to the International Commission on Large Dams, there are 22,000 large dams in China, which account for 46% of the reported dams in the world [85]. However, the studies on GHG emissions are limited in Chinese reservoirs [86]. To evaluate the GHG emissions from the Chinese reservoirs, we could firstly choose the representative reservoirs in the 13 largest hydroelectricity bases in China, measure the 4 main pathways for GHGs to estimate the total GHG emissions from these reservoirs, and estimate the total GHG emissions from Chinese reservoirs based on these results. In the future international negotiations, the results would provide new evidences about whether the development of hydroelectricity could reduce the carbon emission in China.