Contrasting methane emissions from upstream and downstream rivers and their associated subtropical reservoir in eastern China

Subtropical reservoirs are an important source of atmospheric methane (CH4). This study investigated the spatiotemporal variability of bubble and diffusive CH4 emissions from a subtropical reservoir, including its upstream and downstream rivers, in eastern China. There was no obvious seasonal variation in CH4 emissions from the main reservoir, which increased slightly from the first half year to the next half year. In the upstream river, CH4 emissions were low from February to June and fluctuated widely from July to January due to bubble activity. In the downstream river, CH4 emissions were lowest in February, which was possibly influenced by the low streamflow rate from the reservoir (275 m3 s−1) and a short period of mixing. There was spatial variability in CH4 emissions, where fluxes were highest from the upstream river (3.65 ± 3.24 mg CH4 m−2 h−1) and lowest from the main reservoir (0.082 ± 0.061 mg CH4 m−2 h−1), and emissions from the downstream river were 0.49 ± 0.20 mg CH4 m−2 h−1. Inflow rivers are hot spots in bubble CH4 emissions that should be examined using field-sampling strategies. This study will improve the accuracy of current and future estimations of CH4 emissions from hydroelectric systems and will help guide mitigation strategies for greenhouse gas emissions.

emissions were greater in summer than in other seasons at the Three Gorges Reservoir and were regulated by temperature, DO, and water velocity 2 , whereas they were only regulated by temperature at three lakes (Följesjön, Erssjön, and Skottenesjön) in southwest Sweden 20 . Analysis of these differences in effects of environmental factors on spatiotemporal variability in CH 4 emissions from reservoirs may result in more accurate estimates of the total CH 4 emissions than previously determined.
Emissions occur from rivers downstream of reservoirs, due to degassing fluxes at turbines and spillways. A large quantity of CH 4 emits in the downstream river when the hypolimnion water passes through turbines and spillways because of the differences in temperature and pressure. The rapid stream of water increases the water current velocity, which enhances the gas transfer velocity at the air-water interface and improves downstream CH 4 emission flux 22 . 50% of the total CH 4 emissions recorded downstream from the Balbina Reservoir in Brazil 18 represented approximately 30% of the total greenhouse gas emissions from the eight reservoirs in the dry tropical biome region of the country 23 , whereas downstream emissions accounted for 10% of the total CH 4 emissions from the Nam Theun 2 Reservoir in Laos 24 .
In this study, we compared CH 4 emissions from a reservoir with sites upstream and downstream to quantify spatial variations in CH 4 emissions to ensure a more accurate estimation of CH 4 emissions from hydroelectric reservoir systems. Specifically, we tested the hypothesis that upstream and downstream CH 4 emissions are greater than from a reservoir.
Temporal variation in diffusive CH 4 emissions. CH 4 emissions from the upstream river (NW) were low from February to June, but increased and fluctuated from July to January (Fig. 2). Furthermore, on a monthly scale, mean diffusive CH 4 fluxes during the sampling period were similar and generally constant over time among the three areas of the main reservoir; however, fluxes peaked in the southwest (SW) lake on 1 August (DOY: 213) and 8 February (DOY: 39; Fig. 2). On a seasonal scale, there were similar seasonal patterns in CH 4 fluxes among the three areas of the reservoir, where they were lowest in the spring and highest in the autumn (see Supplementary Fig. S4). Mean CH 4 fluxes on the northeast (NE), SW, and southeast (SE) lakes in the next half year were 1.72, 1.54, 1.57 times as many as those in the first half of year, respectively. In addition, there was some temporal variation in mean CH 4 emissions downstream of the reservoir (DR), where it was highest in December 2014 and lowest in February 2015; otherwise, emissions were generally constant (Fig. 2). spatial variation in CH 4 emissions. Mean flux in CH 4 emissions from the upstream river was 3.65 ± 3.24 mg CH 4 m −2 h −1 , whereas mean bubble CH 4 flux (NW-B) was 2.73 ± 2.02 mg CH 4 m −2 h −1 and diffusive CH 4 flux (NW-D) was 0.92 ± 1.22 mg CH 4 m −2 h −1 (Fig. 3). Although there were no bubble CH 4 emissions in the reservoir or the downstream river, the mean diffusive CH 4 emission flux in the reservoir was 0.082 ± 0.061 mg CH 4 m −2 h −1 (NE: 0.076 ± 0.049 mg CH 4 m −2 h −1 , SW: 0.106 ± 0.083 mg CH 4 m −2 h −1 , and SE: 0.064 ± 0.034 mg CH 4 m −2 h −1 ), which was lower than in the downstream river, where it was 0.49 ± 0.20 mg m −2 h −1 (Fig. 3). Mean diffusive CH 4 emissions from the upstream and downstream rivers were higher than those from the reservoir by a factor of 11 and 6, respectively (Fig. 3).
There was no significant difference in mean CH 4 emissions from the marginal to pelagic zones among the three sampling areas of the reservoir (see Supplementary Fig. S5C-E); however, the mean CH 4 emissions from the nearest sampling-point in the downstream river (DRP1: 0.78 ± 0.44 mg CH 4 m −2 h −1 ) were significantly www.nature.com/scientificreports www.nature.com/scientificreports/ higher than those from the second nearest sampling-point (DRP2: 0.34 ± 0.30 mg CH 4 m −2 h −1 ; P < 0.001; see Supplementary Fig. S5A), and the average CH 4 emissions from the pelagic zones of the upstream river were significantly higher than those from the marginal zone (see Supplementary Fig. S5B).
Effects of temperature and wind speed on CH 4 emissions. CH 4 flux from the reservoir was positively correlated with wind speed and air-water temperature difference, whereas CH 4 flux from the downstream river was positively correlated with air-water temperature difference (see Supplementary Tables S1 and S2).

Discussion
Comparison of CH 4 emissions with other reservoirs. Average CH 4 emissions from the main reservoir (0.082 ± 0.061 mg CH 4 m −2 h −1 ) are lower than those from the other temperate and subtropical reservoirs listed in Table 1, except for Douglas Lake, which is presumably due to the deep, oxic conditions and clean water quality in Xin'anjiang Reservoir 25,26 . The mean CH 4 emissions from the upstream river in the study are comparable to that from Three Gorges Reservoir (2.72 mg CH 4 m −2 h −1 ), which is one order of magnitude greater than that from Eguzon Reservoir (0.24 mg CH 4 m −2 h −1 ), but significantly lower than those from Australian reservoirs and an agriculturally impacted reservoir in the United States, due to the differences in bubble activity ( Table 1). The heterogeneity, specifically, differences in ebullition frequency and ebullition magnitudes, contribute to the variability in average CH 4 fluxes observed among the reservoirs 12    www.nature.com/scientificreports www.nature.com/scientificreports/ In regard to the downstream river of the reservoir, it has comparable CH 4 emissions levels with the other listed reservoirs in Table 1. seasonal variation in CH 4 emissions. In the upstream river, CH 4 emissions in autumn and winter are higher than those in spring and summer (Fig. 4), due to the differences in the frequency of bubbles (22.6% versus 8%), but the differences do not reach a significant level (p > 0.05) after performing a one-way ANOVA test. However, the results measured by the bubble traps indicate that the bubble CH 4 emissions in summer and autumn are significantly higher than those in spring (Fig. 1). One of the major differences between the two methods is the duration of the measurement. The measurements using bubble traps were performed over 20-33-h periods, whereas chamber measurements were conducted for 20-30 min only. The floating chambers captured both ebullition and diffusive gas emissions 27 , whereas only CH 4 ebullition fluxes were collected using bubble traps 8 . However, the average ebullitive CH 4 flux (22.62 ± 15.1 mg CH 4 m −2 h −1 ) measured using bubble traps was approximately 5 times higher than that measured using floating chambers (3.65 ± 3.24 mg CH 4 m −2 h −1 ). These differences can be explained by the sudden release of bubbles on these rare occasions, which reveals strong spatiotemporal heterogeneities of the ebullition process because ebullition is highly sporadic and occurs during a very short period of time 7 . The measurements using floating chambers are conducted over a short period of time and a small surface might lead to an underestimation of this emission pathway if hot spots and hot moments are missed during the deployment of the chambers. Such a phenomenon is strongly smoothed when using bubble traps over longer periods of time than the typical floating chamber deployment time (20-33 h versus 20-30 min) 30 .  www.nature.com/scientificreports www.nature.com/scientificreports/ Another explanation for the differences in CH 4 emissions from the upstream river is that they were measured in different years (2014-2015 versus 2016-2017). Admittedly, interannual variability in upstream CH 4 emissions presumably caused unnecessary errors. However, if the average CH 4 flux is calculated only from these bubble-captured chambers in the NW transect in 2014 and 2015, it is 16.83 ± 12.48 mg CH 4 m −2 h −1 , which is approximately 25% less than that measured by the bubble traps (22.62 ± 15.1 mg CH 4 m −2 h −1 ) in 2016 and 2017. Although both the interannual variability and different methods contributed to variances, differences remained in CH 4 emissions when the diffusive and ebullitive CH 4 fluxes were synchronously measured due to episodic bubbles. Nevertheless, the partitioning of bubble and diffusive CH 4 emissions is an uncertainty in this study.
We observed that mean CH 4 emissions from the reservoir in the second half of the year were higher than in the first half of the year (Fig. 2); this was due, in part, to seasonal hydrological dynamics. Hydrology mediates many biogeochemical processes, such as O 2 concentration and thermal stratification, in aquatic systems. Zhang et al. (2015) found that oxycline and thermocline progressively sank in Xin'anjiang Reservoir in the second half of the year 25 . Vertical transport of CH 4 in the water column is typically limited by slow rates of diffusion through the thermocline or oxycline 31 , and thermal and DO stratification typically become weaker in the second half of year, presumably resulting in increases in CH 4 flux at the air-water interface.
Xin'anjiang Reservoir is a thermal stratification lake characterised by a short mixing period in February and March 25 . However, the vertical distributions of CH 4 and O 2 concentrations, and temperature were not measured in this study, which failed to illuminate the temporal variability in CH 4 emissions. Lake overturn is a hot moment that exhibits disproportionately high CH 4 emissions and CH 4 oxidation 32 . Many studies indicate that CH 4 storage sharply decreases during seasonal overturn periods [32][33][34][35] , emitting 12-46% of the total CH 4 to the atmosphere, whereas the remainder (54-88%) is consumed by methane-oxidizing bacteria 32,34,35 . Although a minor proportion of the storage CH 4 was emitted to the atmosphere, the contribution to the annual diffusive CH 4 emissions was still great 32,34,35 , and even extremely diffusive CH 4 fluxes occurred 33 . However, CH 4 emissions from the main reservoir did not show a pulse in February and May (Fig. 2), presumably because the low measurement frequency in our study did not capture the CH 4 emission peaks. Thermal stratification and its impact on CH 4 emissions is important to understanding the mechanisms of the spatial and temporal variability of CH 4 emissions from reservoirs, which should be examined in the near future.
We recorded a clear peak in CH 4 emissions (0.25 ± 0.15 mg CH 4 m −2 h −1 ) on 1 August (DOY: 213) in the SW lake (Fig. 2), which was presumably due to the fluxes from the two marginal sampling points (SWP1 and SWP2) of 0.47 ± 0.11 mg CH 4 m −2 h −1 and 0.33 ± 0.061 mg CH 4 m −2 h −1 , respectively (see Supplementary Table S7). The high CH 4 fluxes from the marginal zone may be attributed to the decomposition of vegetation in the littoral zone when the water level increased to its highest point (104.4 m) in July (see Supplementary Fig. S2). It is likely that the gentle slopes that had adequate levels of soil on the banks of the SW transect permitted the growth of vegetation in the littoral zone during the spring when water levels were low, whereas the banks of the NE and SE lakes were steep and rocky and presumably less well vegetated. Similar peaks in CH 4 emissions have also been reported from littoral zones of the Miyun and Three Gorges Reservoirs 5,36 .
We recorded another peak, albeit low (0.16 ± 0.097 mg CH 4 m −2 h −1 ) on 8 February (DOY: 39) in the SW lake (Fig. 2), which was caused by strong winds. Gas samples were only collected from three of the five sampling points due to unstable safety conditions on the surface of the reservoir. Mean CH 4 fluxes were 0.23 and 0.20 mg CH 4 m −2 h −1 at SWD2 and SWD4 (see Materials and Methods), respectively, when the wind speed reached 8-10 m s −1 , whereas the lowest CH 4 flux at SWD5 (0.049 mg CH 4 m −2 h −1 ) occurred in the central area of the reservoir due to the low wind speed (2.63 m s −1 ). Many studies support the opinion that CH 4 emissions from water surfaces can be enhanced by strong wind speeds 15,22,33,37 .
Downstream CH 4 emissions (including degassing at the turbines) have been found to be proportional to streamflow 19 . It is impossible to calculate the degassing emissions from the turbines at Xing'anjiang Dam based on the differences in CH 4 concentrations between the water intake and water outlet below the dam because access is forbidden 500 m upstream and downstream of the dam due to safety concerns. However, measurements of CH 4 emissions at four distances downstream of the dam, taken 13 times in 2015 (see Supplemental Table S9), were found to be at their lowest (0.17 ± 0.11 mg CH 4 m −2 h −1 ) in February, which is presumably a result of a low discharge flow rate (275 m 3 s −1 ). Another possible explanation for the low flux is related to the lake overturn phenomenon in February 25 . Most of the CH 4 stored in the hypolimnion is oxidized or released to the atmosphere during overturn periods [32][33][34][35] , and a very small fraction of the original quantity of CH 4 remains in the water column 32 ; thus, a low CH 4 flux level was measured in the downstream river in February (Fig. 2). spatial variation in CH 4 emissions. Upstream CH 4 emissions are hot spots because they exhibit disproportionately high ebullitive CH 4 emissions relative to the surrounding matrix 38 . Upstream river CH 4 emission dynamics are predominantly influenced by bubbles since the peaks in the CH 4 emissions flux (Fig. 2) are driven by bubbles (see Supplementary Table S4). In contrast to other studies 21,24 , we found that bubbles occurred in the deep-water zone (>10 m) rather than in the shallow zone (<5 m), and we suggest that the high ebullitive CH 4 emissions from deep water zones are related to heterogeneous sediment accumulation 12,13 because little or no sediment accumulates along reservoir margins 39 .
The average CH 4 emission rate at the upstream site (NW) was one to 2 orders of magnitude greater than the other sites (Figs 2 and 3), highlighting the importance of identifying ebullition hot spots to improve total emissions estimates 27 . The results supported our hypothesis that CH 4 emissions are higher in rivers upstream and downstream of the reservoir than in the main reservoir (Fig. 3), where high CH 4 emissions from the upstream river were mediated by bubbles (Figs 1 and 3, see Supplementary Table S4). The CH 4 in the gas bubbles can escape oxidation during transport through the water column as CH 4 moves faster through the water column by ebullition than by diffusion 40 (Table 2), which may be attributable to the fact that water slows down in these areas and sediments have higher chances for deposition 14 . Sediment accumulation rates are positively correlated to the areal organic carbon burial rates 39 , and rapid burial of fresh sediments and organic matter made upstream sites more carbon rich and prime for CH 4 production by anaerobic metabolism compared to other parts of the reservoirs 12,15,27,37 , as CH 4 production in reservoirs is strongly driven by organic carbon availability 41 . Thus, ebullitive CH 4 emissions are often reported to be exponentially increased with corresponding sediment accumulation rates 14,42 . The upstream reaches of Xin'anjiang Reservoir directly receive the catchment and stream inflow of industrial and domestic pollution 43 , which presumably fostered high rates of sediment CH 4 production in the upstream rivers of the reservoir, causing ebullition zones to subsequently appear 27,29 . Moreover, ebullition rates tend to be highest in shallow areas because short water residence times limit the dissolution of CH 4 -rich bubbles released from the sediment 44 . The upstream river is the shallowest area compared with other regions (see Supplementary Table S3), which is beneficial for bubbles transport from the sediment to the atmosphere because of the small proportion of dissolved gas bubbles during ascent 15,27,37 . Additionally, CH 4 imported from the Xin'anjiang catchment may further contribute to the observed pattern at river inflow areas.
We recorded higher CH 4 emissions from the downstream river than from the surface of the reservoir adjacent to the dam (Fig. 2) that had presumably been released from dissolved CH 4 in the hypolimnion layer of the reservoir 17 because water inlets of turbines located in the hypolimnion layer (26-37 m under water surface) 43 and the discharged water derived from the hypolimnion layer almost year round (except February, due to mixing periods). The water adjacent to the dam is thermally stratified, where water in the warmer, upper layer (epilimnion <33 m) is in contact with the atmosphere and is more oxygen-rich, whereas the deeper, colder layer (hypolimnion) contains relatively low levels of O 2 concentration 25 . We suggest that CH 4 produced in the reservoir is easily stored in the hypolimnion 45 , and the release of dissolved CH 4 to the atmosphere occurs due to differences in pressure, temperature, and turbulence when water passes through the turbines and spillways 19 . Water passing through the turbines and spillways is drawn from the hypolimnion, and downstream CH 4 emissions are released under decreased pressure below the dam 19 .
The explanation for the low CH 4 emissions from the main reservoir is that the deep, oxic waterbody slows emissions by offering more options for CH 4 oxidation. Water depths of the sampling points range from 10-69 m, except for those on the margin (Table S3), and it is possible that such reservoir depths increase the possibility of oxidization for diffusive CH 4 molecules. Moreover, Zhang et al. (2015) reported that the DO concentration never fell below 2 mg/L, the critical value for anoxia, in Xin'anjiang Reservoir 25 . The lack of an anoxic layer permits the oxidization of dissolved CH 4 under aerobic conditions by methanotrophic bacteria 27 . Furthermore, biomass clearing before flooding limited the availability of organic carbon 26,43 , which is important for CH 4 production in sediments 41 . Chlorophyll a is a significant predictor of CH 4 emissions from reservoir water surfaces 1,37 , and Xin'anjiang Reservoir is presently in an oligotrophic state, with a low concentration of chlorophyll a (1-3 μg L −1 ) 26 , which limits CH 4 emissions from the reservoir. Moreover, the dendritic shape of Xin'anjiang Reservoir facilitates the deposition of allochthonous organic carbon in the sediment of the NW lake (see Supplementary Fig. S1) 46 , and limited fresh sediments are deposited in the main reservoir.
Mitigation strategies for CH 4 emissions. Management strategies should increase CH 4 oxidation in the sediments and water columns and decrease CH 4 production, ebullition, and degassing emission at the dam to mitigate CH 4 emissions from reservoirs. Extremely allochthonous organic material and organic carbon burial stimulated ebullition in the upstream rivers and river deltas 12,14 ; therefore, periodical dredge campaigns 27 , reducing watershed soil erosion 14 and nutrient input 47 , can efficiently reduce ebullitive CH 4 emissions. Moreover, the location of spillways and turbines have an impact on CH 4 emissions from reservoirs 27 .
Previous studies have shown that extreme CH 4 ebullitive emissions are ultimately attributable to very high sedimentation rates 14 , as well as exhibiting an exponentially increasing relationship between CH 4 ebullitive location observations ref

Three Gorges Reservoir, China
Upstream, reservoir tail waters and tributary sites had higher CH 4 fluxes than the mainstream of the reservoir. 3 Lake Kariba, Zambia/Zimbabwe Higher fluxes in river deltas (~10 3 mg CH 4 m −2 d −1 ) than nonriver bay (less than 100 mg CH 4 m −2 d −1 ) due to the high ebullition frequency and ebullition magnitudes. 27 Little Nerang Dam, Lake Wivenhoe, Lake Baroon, Australia CH 4 saturation was higher in inflow zones than in the main body. 27 William H. Harsha Lake, USA Extreme high CH 4 emission (mean: 3137 ± 660 mg CH 4 m −2 d −1 ) at the most upstream site; 1 to 2 order of magnitude greater than the other sites. 27 Glod Creek Reservoir, Australia Highest CH 4 water-air fluxes were found at the main water inflow areas of the reservoir. 28 Little Nerang Dam, Australia 1.8-7.0% of the upstream surface area called "ebullition zone"; 97% of the total methane occurred in the ebullition zones. 29 Chapéu D'Uvas, Curuá-Una, Furnas, Brazil Elevated pCH 4 and CH 4 concentrations in river inflow areas and decreasing values toward the dam; River inflows are hot spots of diffusive C gas flux. 37 Table 2. Some examples of studies reporting high methane emissions from the upstream inflow areas of reservoir.
www.nature.com/scientificreports www.nature.com/scientificreports/ emissions and the sediment accumulated rates in the 6 small reservoirs of the Saar River 42 . The mechanism is characterised by deeper sediment layers contributing to CH 4 formation 42 , and the deeply accumulated CH 4 causes supersaturation and consequent bubble formation and release 14 . In the study, inflow rivers are ebullition hot spots, thus policymakers should take effective measures to control substantial CH 4 emissions. The sediment is dredged periodically to reduce deposited organic matter, which presumably decreases the magnitude of ebullitive CH 4 emissions efficiently, although carbon leakage occurs during the process 27 .
Another practical measure is to prevent the excessive input of nutrients and pollution to the reservoir 30,47,48 , which would reduce the available organic carbon for CH 4 production 47 . Cage culture is an important nutrient input, which enhances N, P, and TOC accumulation in the sediments of the lacustrine zone 48 . Moreover, the NW lake received more soil erosion, sewage input, and industrial pollution from the upstream rivers in Anhui Province. In response, authorities have taken measures to decrease the inputs of all types of pollution, such as cage culture prohibition and inter-provincial ecological compensation, which improved the water quality in the upstream river from a eutrophic to mesotrophic state, presumably decreasing CH 4 production and emissions from the reservoir 37 .
Dam design is also important for CH 4 emissions, especially the location of water intakes. CH 4 concentrations are higher in the hypolimnion than in the epilimnion during thermal stratification periods 19 . The degassing that occurs as hypolimnion water is routed through a dam accounts for a large fraction (>50%) of the total CH 4 emissions in some Amazon tropical reservoirs 17,18 . However, if turbine intakes are located in the upper layer of a dam, shallow waters will be withdrawn during thermal stratification to avoid substantial CH 4 degassing from the CH 4 -rich water in the hypolimnion 19 , for example, only 0.8% from Harsha Lake 27 . Moreover, a significant increase in CH 4 emissions was reported 3 km upstream from Nam Theum 2 Dam due to the artificial mixing induced by water intakes 33 , and CH 4 -rich water from the reservoir's hypolimnion reached the surface and resulted in a high CH 4 diffusive flux. Therefore, the water intake in the hypolimnion not only increased the degassing flux at the dam but also risked enhancing the CH 4 diffusive flux upstream of the dam.
In summary, upstream rivers are hot spots in bubble CH 4 emissions, significantly contributing to the total CH 4 emissions from hydroelectric reservoir systems. If upstream sites are ignored in field-sampling strategies, entire-system CH 4 emissions will be underestimated. CH 4 emissions from a main reservoir are lower than that from a downstream river. Capturing the spatial heterogeneity of CH 4 emissions is vital to estimating the total CH 4 emissions in a hydroelectric system. Seasonal variation in CH 4 emissions exhibited a high value in autumn and winter and a low value in spring and summer. A thorough investigation should be conducted for the entire reservoir region over a long period because bubbles are episodic and diffusive CH 4 emission flux exhibits a strong spatiotemporal variability.

Materials and Methods
study sites. Xin'anjiang Reservoir (118°42′-118°59′E, 29°28′-29°58′N) is located in China's north subtropical zone. The mean annual air temperature, precipitation, and evaporation are 17.7 °C, 2015.1 mm, and 712.9 mm, respectively (see Supplementary Fig. S2). Constructed in 1959, the reservoir has a water surface area of 580 km 2 and mean depth of 37 m, with a capacity of approximately 1.78 × 10 10 m 3 43 , and an annual average inflow and outflow of 9.4 × 10 9 m 3 and 9.1 × 10 9 m 3 , respectively. Water retention time is approximately 2 years, and in 2015, the water level fluctuated between 98 and 104 m above elevation (see Supplementary Fig. S2). According to China's surface water classification standards, the water quality of Xin'anjiang Reservoir is grade I, serving as an important water source in eastern China that presently provides drinking water.
The reservoir consists of a series of connected lakes in all cardinal directions around a central lake that serves as the main waterbody (see Supplementary Fig. S1). The watercourse of the northwest lake is the dominant source of upstream inflow, contributing 60-80% of the total inflow. The downstream river is the watercourse below Xin'anjiang Dam.
The four sub-lakes and downstream river were sampled at points along transects (see Supplementary Fig. S1). The NW lake transect (118°43′04″E, 29°44′03″N), located in the main upstream inflow inlet, has a width of 0.3 km and three sampling points extending 10, 50, and 120 m (NWP1, NWP2, and NWP3, respectively) from the southern bank marginal zone to the pelagic zone, whereas the NE (119°03′03″E, 29°38′44″N), SW (118°44′39″E, 29°28′18″N), and SE (118°45′20″E, 29°28′39″N) lake transects are located in the open water and have five sampling points (P1 to P5) extending from the marginal to pelagic zones (Table S3). Four sampling points in the downstream river below the dam are located 0.35, 1, 4, and 7 km from Xin'anjiang Dam (DRP1, DRP2, DRP3, and DRP4, respectively). CH 4 flux measurement. Floating static chambers were used to collect gas samples at all sampling points between 08:30 and 11:30 hrs, monthly from December 2014 to December 2015, and bubble traps were used to collect bubbles from the upstream river from August 2016 to November 2017, where samples were collected once or twice per month, except November 2016, and January and February 2017. Air and water temperatures were measured using an alcohol thermometer, and wind speed in the field was measured using an anemometer (Kestrel 1000, Nielsen-Kellerman Co., USA).
Flux of diffusive CH 4 emissions was collected using floating static chambers and analysed by gas chromatograph. Three floating static chambers (basal area of 0.29 m 2 and volume of 0.117 m 3 ) at each sampling point comprised a non-covered plastic box wrapped in light-reflecting and heatproof materials to minimize internal temperature variation, with plastic foam collars fixed to opposite sides. The headspace height inside the chamber was approximately 35 cm. A silicone tube (0.6 and 0.4 cm outer and inner diameters, respectively) was inserted into the upper central side of the chamber to collect gas samples that were then dried to prevent biological reactions in plexiglass tubes filled with calcium chloride (anhydrous, analytical reagent). Another silicone tube was inserted into the upper corner of the chamber to maintain a balance in air pressure between the inside and outside www.nature.com/scientificreports www.nature.com/scientificreports/ of the chamber. Static chambers drifted freely behind a boat to reduce measurement bias 49 . Samples of gas were collected from the static chamber in air-sampling bags (0.5 L, Hedetech, Dalian, China) four times every 7 min over a 21-min period using a hand-driven pump (NMP830KNDC, KNF Group, Freiburg, Germany) and were stored until analysis 2 . The air-sampling bags made of aluminium are suitable for gas storage for 7 days and do not absorb or react with CH 4 . Leakage and memory effects of the air-sampling bags were tested in earlier experiments.
We placed 16-26 bubble traps 10-15 m apart in a river crossing rope in the upstream river, where water depth ranged from 5-25 m. The traps consisted of an inverted 30-cm diameter circular funnel fixed to the neck of a 0.56-L plastic bottle, and an additional skirt (50-cm diameter) was fixed to the funnel aperture to enlarge the area over which bubbles were collected 8 . Each funnel was stabilized with three equally sized weights to ensure no tiny bubbles remained in the traps at the initial stage. Trapped gas bubbles liberated from water were collected in the bottles after 24 hours, and then the remaining volume of water was measured to calculate the volume of liberated gas bubbles. The trapped gas was diluted 1000 times by injecting 1 × 10 −3 -L of trapped gas into 1-or 0.5-L gas bags that had been filled with N 2 to facilitate analysis of CH 4 concentration by gas chromatography. Trapped gas within these bags was analysed within 3 days using a gas chromatograph (Agilent 7890 A, Agilent Technologies, Santa Clara, USA) equipped with a flame ionization detector (FID). The oven, injector, and detector temperatures were set at 70, 25, and 200 °C, respectively. Standard mixed gas (CH 4 : 1.83 ppm, provided by the China National Research Centre for Certified Reference Materials, Beijing) was used to quantify the CH 4 concentration in one of every 10 samples, and the coefficient of variation of CH 4 concentration in the replicated samples was <1%.
The increasing rate of gas concentration (dc/dt) within the static chamber was calculated as the slope of the linear regression of the gas concentration versus time. Diffusion chambers collect diffusive emissions as well as ebullitive emissions if they are present. Therefore, if the slope of the linear regression of the gas concentration in the chamber versus time was linear, with R 2 > 0.9, then the chamber was assumed to collect only diffusive emissions. If R 2 < 0.9, then the chamber was assumed to collect total (diffusive + ebullitive) emissions 30  where C CH4 is the CH 4 concentration (μL L −1 ), V is the accumulated headspace gas volume (L), M is the molar weight of CH 4 (16.04 g mol −1 ), A f is the funnel area (0.14 m 2 ), t is the measurement duration (h), and V m is the molar volume of gas at room temperature (22.4 L mol −1 ) 8 . The ebullition rate (ER; mL m −2 h −1 ), which reflects the volume rate of released accumulated bubbles, is calculated as (Eq. 3).
where the parameters V, A f , and t are provided in Eq. (2). statistical analysis. The flux in CH 4 emissions data that did not meet the test for normality (Kolmogorov-Smirnov) were transformed to trigonometric or logarithmic functions prior to testing for seasonal and spatial variability using one-way analysis of variance (ANOVA) and Tukey's HSD test. Data were analysed using the SPSS statistical package (v. 18.0, Chicago, IL, USA).