Carbon Dioxide Emissions from the Littoral Zone of a Chinese Reservoir

: The continuous increase in the number of reservoirs globally has raised important questions about the environmental impact of their greenhouse gases emissions. In particular, the littoral zone may be a hotspot for production of greenhouse gases. We investigated the spatiotemporal variation of CO 2 ﬂux at the littoral zone of a Chinese reservoir along a wet-to-dry transect from permanently ﬂooded land, seasonally ﬂooded land to non-ﬂooded dry land, using the static dark chamber technique. The mean total CO 2 emission was 346 mg m − 2 h − 1 and the rate varied signiﬁcantly by water levels, months and time of day. The spatiotemporal variation of ﬂux was highly correlated with biomass, temperature and water level. Flooding could play a positive role in carbon balance if water recession occurs at the time when carbon gains associated with plant growth overcomes the carbon loss of ecosystem. The overall carbon balance was analysed using cumulative greenhouse gases ﬂuxes and biomass, bringing the data of the present study alongside previously published, simultaneously measured CH 4 and N 2 O ﬂuxes. For the growing season, 12.8 g C m − 2 was absorbed by the littoral zone. Taking CH 4 and N 2 O into the calculation showed that permanently ﬂooded sites were a source of greenhouse gases, rather than a sink. Our study emphasises how water level ﬂuctuation inﬂuenced CO 2 , CH 4 and N 2 O in different ways, which greatly affected the spatiotemporal variation and emission rate of greenhouse gases from the littoral zone.


Introduction
Greenhouse gas (GHG) emissions from reservoirs have attracted much attention, not only because of the substantial contribution they already make to global GHG budgets but also because large reservoirs are still being created in many parts of the world, especially Southeast Asia, South America and Africa [1,2].When reservoirs are designed for power generation they are often assumed to be 'clean' when the alternative to hydropower involves combustion of fossil fuels or the use of nuclear power stations.Yet even this assumption has been questioned [3][4][5][6], and, in the absence of data on the emissions of CH 4 , N 2 O and CO 2 , there are few cases where the full environmental costs and benefits have been investigated.
Years ago, working with limited data, the global CO 2 emissions from reservoirs were estimated to be 273 Tg C year −1 , which was then equivalent to 4% of other anthropogenic emissions [3].
Water 2017, 9, 539 2 of 13 However, the latest global estimate was much lower when including more data and environments.Deemer et al. [4] reports 36.8Tg C year −1 while the number from Li & Zhang [5] reported here is 82 Tg C year −1 .Those researchers also emphasized the importance of drawdown and downstream emissions which can represent up to 42% of total emissions, as well the large degree of uncertainty caused by spatial and temporal variation [4][5][6][7].Indeed, much uncertainty remains.For example, one recent study in CO 2 concentrations and fluxes of five tropical reservoirs pointed out that ignoring the spatial variability can lead to more than 25% error in the total system flux [8].CO 2 emission are likely to be underestimated by approximately 40% when ignoring the diel emission difference and frequent synoptic weather events which may induce CO 2 emission pulses at longer time scales [7].
Of particular interest is the intensive and highly varied greenhouse gas emission around the margins of reservoirs.As a characteristic of the littoral zone, flooding influences a plethora of environmental variables, which include vegetation change/succession, hydrostatic pressure, soil aeration, nutrients, and soil/sediment temperatures, all of which add to the variation in the greenhouse gas production and emission rate [9][10][11][12][13][14].During the flooding phase, higher variability of CO 2 emission was observed within the various sections of each temporary zone (i.e., inundated, emerged-unvegetated and emerged-vegetated) than among the zones [12].Exposure of sediment could activate oxidase and hydrolase (two magnitudes higher), leading to more greenhouse gases emission [15].When the water table decreased from 5 cm aboveground to 5 cm belowground, the total soil respiration of the growing season increased from 720 g m −2 to 1490 g m −2 [16].Although the influences of water level fluctuation on greenhouse gases emission have been observed, the paucity of studies on these various effects constrain the overall understanding on greenhouse gases emission from drawdown areas.
Elsewhere, we have reported the emissions of CH 4 and N 2 O from the Miyun Reservoir, and we demonstrated that this littoral zone can be regarded as a hot spot for emissions [17,18].In the present paper, we evaluate (i) the spatiotemporal variation of CO 2 emissions from the littoral zone; and (ii) the influence of environmental factors, especially flooding.Finally, we conclude our analysis by considering the complete greenhouse gas balance of the littoral zone in which the contributions of CH 4 and N 2 O are considered alongside CO 2 .

Study Area and Sampling Sites
The research was carried out at the littoral zone of the Miyun Reservoir (40 • 29 N, 116 • 50 E), Beijing, China.Its catchment is classified as warm temperate semi-humid monsoonal climate, having a hot summer and a cold winter.The growing season is from April to November.Precipitation is unequally distributed across the typical year (80% in summer) [19].The annual water level difference is around 1 to 5 m [20].
Along the transect from water to upland, the sampling sites are grouped at 5 different heights (Figure 1), i.e., (1) DW: deep water area; (2) SW: shallow water area; (3) SF: seasonally flooded area; (4) SFC: 'control area' for the seasonally flooded area, which had similar vegetation and soil moisture as site SF before SF was flooded, but escaped the flood in August and September because of its 1-m-higher elevation; (5) NF: non-flooded area, which usually floods once per several years but was not flooded in the sampling year.Three plots supporting somewhat different vegetation types were selected within each height band, for the species details see Supplementary Materials Table S1.There were 15 plots in total, four replicates in each plot.
The study area and the experimental design of the present research was exactly the same as in the previous CH 4 and N 2 O study [17,18].     2 and S1.
Table 1.Physicochemical properties (Mean ± SE (standard error)) of soil/sediment of each site.DW: deep water site; SW: shallow water site; SF: seasonally flooded site; SFC: 'control site' for the seasonally flooded site; NF: non-flooded site.A, B and C denote samples from different vegetation types within each height band, species details see Table S1.

Site
Bulk Density (g cm

CO2 Flux and Environmental Factors Measurements
CO2 flux was observed using the dark chamber and gas chromatography techniques (for details see [17,18]).The gas samples were repeatedly taken six times in the year to cover the different seasons (November 2011, May, July, August, September and October 2012) and covering the transition in and out of the flooding season.Also to capture diurnal variation, the plots were repeatedly sampled seven times per day (local time: 6, 9, 12, 15, 18, 21 and 24 h).To determine the effect of aboveground vegetation on flux, one more flux measurement was taken at 9 a.m. the following day (after seven times sampling for diurnal variation), with the aboveground plant material removed.Chambers were Experimental design [17,18].WL: water level.The difference between high WL and low WL was caused by summer flooding.m H indicates meters in horizontal; m V indicates meters in vertical.The sites are grouped at different heights.DW: deep water site; SW: shallow water site; SF: seasonally flooded site; SFC: 'control site' for the seasonally flooded site; NF: non-flooded site.A, B and C denote samples from different vegetation types within each height band.For more details on water depth and other environmental parameters, see Table 1, Figures 2 and S1.
Table 1.Physicochemical properties (Mean ± SE (standard error)) of soil/sediment of each site.DW: deep water site; SW: shallow water site; SF: seasonally flooded site; SFC: 'control site' for the seasonally flooded site; NF: non-flooded site.A, B and C denote samples from different vegetation types within each height band, species details see Table S1.

CO 2 Flux and Environmental Factors Measurements
CO 2 flux was observed using the dark chamber and gas chromatography techniques (for details see [17,18]).The gas samples were repeatedly taken six times in the year to cover the different seasons (November 2011, May, July, August, September and October 2012) and covering the transition in and out of the flooding season.Also to capture diurnal variation, the plots were repeatedly sampled seven times per day (local time: 6, 9, 12, 15, 18, 21 and 24 h).To determine the effect of aboveground vegetation on flux, one more flux measurement was taken at 9 a.m. the following day (after seven times sampling for diurnal variation), with the aboveground plant material removed.Chambers were reset into new positions near the old positions each sampling month.For more details on the water depth and other environmental parameters, see Figure S1 and Table 1.
In order to calculate the cumulative carbon balance, the belowground biomass of every replicate in the chamber of SW, SF and NF was weighed after rinsing and drying at 80 • C to constant mass.The belowground biomass of DW was not determined directly because of manipulation difficulty.To protect the sampling area from serious interference, root biomass was collected just one time in November 2011.The observed average root/shoot ratio of 0.41 was used for each area in the total biomass calculation.The aboveground biomass of every replicate in the chamber was weighed every campaign after drying at 80 • C to constant mass.
The air temperature was observed at the start and end of each gas sampling at every plot by means of a digital thermometer (JM624, Jinming, Henan, China).The standing water depth was measured using a metre stick after gas sampling at DW, SW and SF (during flooding).The water level of SF (before and after flooding) and SFC was shallow.Therefore, a 1-m high PVC (Polyvinyl chloride) tube with small side holes was inserted vertically into the soil under the chamber after all monthly gas sampling was complete, allowing two h for the water level to equilibrate before measuring the level.The water level of site NF was calculated according to the elevation measured by a Global Navigation Satellite System receiver (BLH-L90, Daheng International, Hong Kong, China).
Soil samples at site DW, SW, SF and NF were collected manually (0-30 cm) at each replicate location in November 2011, except site SFC in October 2012.Soil NH 4 + , NO 3 − , pH, total carbon, total nitrogen and bulk density were then analysed.Weekly precipitation, daily air temperature and daily water level were accessed online.Wind speed, soil water content and dissolved oxygen in water were recorded in situ, for detailed methods, see Yang et al. [17].

Greenhouse Gases Balance Assessment
To calculate the cumulative greenhouse gas emissions, models were needed for day-to-day flux simulation.Emission models were selected using the AIC (Akaike information criterion) criterium according to Burnham et al. [21].Following a paper which represents a method for NEE (net ecosystem exchange) assessment based on chamber technique, we choose soil apparent respiration (SAR, sum of root respiration and heterotrophic respiration) for evaluating the carbon balance.After model selection, daily SAR, CH 4 and N 2 O emissions were simulated and then used for C balance calculations, combined with daily biomass changes following [22] with Equation (1): where F SAR is the soil apparent respiraton.α is the ratio of heterotrophic respiration to soil apparent respiration, assuming a value of 0.55 [22,23].GWP 100 is Global Warming Potential over 100-year time-span, which was taken as 34 and 298 for CH 4 and N 2 O respectively [24].Biom is the total biomass including both shoot and root.β is the carbon content of biomass, which was taken as 40% [25].
In the simulations, the total biomass growth was represented as a linear interpolation of the biomass data.Daily water depth was calculated using the daily water level and elevation of each area.Soil properties were assumed stable during the simulated season.

Statistical Analysis
Flux differences among the water levels, months and times of day were tested using a three-way ANOVA, and then using LSD (Least Significant Difference) for multiple comparisons.All the analyses above were performed using IBM SPSS Statistics (version 19.0, IBM, New York, NY, USA).

CO 2 Flux and Environmental Influences
The fluxes (total respiration) were significantly different among water levels, months and times of day (Figure 2 and Table 2).The mean CO 2 flux from the littoral zone was 346 mg m −2 h −1 , ranging Water 2017, 9, 539 5 of 13 from −98 to 2274 mg m −2 h −1 .Negative flux was rarely observed (sometimes at DW-A and DW-B, 8 times in all 324 observations).
Spatially, the flux increased along the gradient from the deep water (114 ± 10 mg m −2 h −1 ) to the non-flooded area (723 ± 10 mg m −2 h −1 ), Figure 2. The efflux from the seasonal flooded area (229 ± 10 mg m −2 h −1 ) was much lower than its control area (629 ± 14 mg m −2 h −1 ), Figure 2. Biomass and water level showed a good correlation with variation of the CO 2 flux among water levels, the r 2 of the linear regression being 0.99 and 0.80 respectively.
Temporally, clear monthly and diurnal variations were observed, Figure 2. The flux peaked in August and mid-day respectively.Temperature showed the best correlation with the temporal variation of CO 2 flux.The change patterns of monthly average water level and CO 2 emission of all sites were similar, but linear regressions were not significant (p = 0.06).The variable 'Biomass' could not explain monthly CO 2 variation, the r 2 of linear regression being as low as 0.003 (p = 0.92).

CO2 Flux and Environmental Influences
The fluxes (total respiration) were significantly different among water levels, months and times of day (Figure 2 and Table 2).The mean CO2 flux from the littoral zone was 346 mg m −2 h −1 , ranging from −98 to 2274 mg m −2 h −1 .Negative flux was rarely observed (sometimes at DW-A and DW-B, 8 times in all 324 observations).
Spatially, the flux increased along the gradient from the deep water (114 ± 10 mg m −2 h −1 ) to the non-flooded area (723 ± 10 mg m −2 h −1 ), Figure 2. The efflux from the seasonal flooded area (229 ± 10 mg m −2 h −1 ) was much lower than its control area (629 ± 14 mg m −2 h −1 ), Figure 2. Biomass and water level showed a good correlation with variation of the CO2 flux among water levels, the r 2 of the linear regression being 0.99 and 0.80 respectively.
Temporally, clear monthly and diurnal variations were observed, Figure 2. The flux peaked in August and mid-day respectively.Temperature showed the best correlation with the temporal variation of CO2 flux.The change patterns of monthly average water level and CO2 emission of all sites were similar, but linear regressions were not significant (p = 0.06).The variable 'Biomass' could not explain monthly CO2 variation, the r 2 of linear regression being as low as 0.003 (p = 0.92).To determine the effect of biomass on fluxes and provide data for the assessment of gas balance, fluxes after removal of aboveground plants were measured.Compared to the fluxes with the shoots intact, the differences between sites largely vanished in term of CO 2 emission among the plots in the same water condition (Figure 3).Removal of the aboveground parts demonstrated that the contribution of aboveground vegetation to total respiration was 27-75%, averaged at 52%.
Using the AIC criterium, seven variables went into the best CO 2 model, i.e., shoot biomass, air temperature, water depth, soil pH, soil NH 4 + , soil total carbon and soil total nitrogen.
Water 2017, 9, x FOR PEER REVIEW 6 of 13 To determine the effect of biomass on fluxes and provide data for the assessment of gas balance, fluxes after removal of aboveground plants were measured.Compared to the fluxes with the shoots intact, the differences between sites largely vanished in term of CO2 emission among the plots in the same water condition (Figure 3).Removal of the aboveground parts demonstrated that the contribution of aboveground vegetation to total respiration was 27-75%, averaged at 52%.
Using the AIC criterium, seven variables went into the best CO2 model, i.e., shoot biomass, air temperature, water depth, soil pH, soil NH4 + , soil total carbon and soil total nitrogen.

Greenhouse Gas Balance
For daily greenhouse gas fluxes, several simulation models we examined using the AIC criterium (Table 3).Of all three gases, CO 2 (soil apparent respiration) is the one that showed the strongest correlation with environmental variables (r 2 = 0.54).The correlation between N 2 O emission and environmental variables was weak (r 2 = 0.15).The greenhouse gas balance is presented as cumulative carbon fluxes and balances (Figure 4).Including CH 4 and N 2 O into the consideration of GHG-balance brought noticeable changes in the overall balances, especially for site DW and SW, turning the littoral area from a sink to a source of greenhouse gases.In deep water, the growth of biomass was insufficient to absorb more than a tiny fraction of the observed emissions (Figure 4a).In shallow water conditions (SW) fixation of carbon as biomass offset a substantial part of the emissions, but there was however a net loss of greenhouse gases from the sites (Figure 4b).For the growing season (184 days), the budget of permanent flooded areas DW and SW, in terms of C-equivalent, was 123.3 g C m −2 and 5.1 g C m −2 .In the seasonal flooded area, regrowth of vegetation after water recession accumulated 227 g m −2 in term of new biomass from exposure of the land at the end of October.The increasing in biomass converts SF from being a greenhouse gas source to a sink (the balance changed from 33 to −70 g C m −2 , Figure 4c).In the non-flooded area, the carbon absorbed by the growing vegetation was substantial (note the change of scale of the y-axis) allowing it to be the largest sink of 109.1 g C m −2 , although it also had the largest emission of 225.0 g C m −2 .For the growing season, 12.8 g C m −2 was absorbed on average.

CO2 Flux
The margins of reservoirs and lakes are often overlooked as potential sources and sinks of greenhouse gases.Most researchers have focused on the water body itself (see [26][27][28], disregarding the green zone at the water's edge.For lakes especially, and for reservoirs to a lesser extent, there are many investigations of the CO2 flux from the open water surface, sometimes measured directly using chambers or (rarely) obtained from micrometeorological techniques; but more often, fluxes are estimated from a knowledge of the dissolved CO2 concentration in the water samples [29][30][31][32].Our study helps to redress the balance by providing new observational data from the margins of a reservoir which may be compared to values obtained from open water.In a review of the global literature on lakes, Pace & Prairie (their Table 7.1, [27]) found a range from open water that varied enormously between 21 and 395 mg CO2 m −2 h −1 with a mean of 148 mg CO2 m −2 h −1 , derived from planktonic and sediment respiration.Recently we have compiled a database of the world literature on reservoirs (supplement Table S3): the mean value of 79 studies at pelagic sites was 103 mg CO2 m −2 h −1 .If we take only the Chinese data from the database (n = 10), we have 72 mg CO2 m −2 h −1 as the

CO 2 Flux
The margins of reservoirs and lakes are often overlooked as potential sources and sinks of greenhouse gases.Most researchers have focused on the water body itself (see [26][27][28], disregarding the green zone at the water's edge.For lakes especially, and for reservoirs to a lesser extent, there are many investigations of the CO 2 flux from the open water surface, sometimes measured directly using chambers or (rarely) obtained from micrometeorological techniques; but more often, fluxes are estimated from a knowledge of the dissolved CO 2 concentration in the water samples [29][30][31][32].Our study helps to redress the balance by providing new observational data from the margins of a reservoir which may be compared to values obtained from open water.In a review of the global literature on lakes, Pace & Prairie (their Table 7.1, [27]) found a range from open water that varied enormously between 21 and 395 mg CO 2 m −2 h −1 with a mean of 148 mg CO 2 m −2 h −1 , derived from planktonic and sediment respiration.Recently we have compiled a database of the world literature on reservoirs (supplement Table S3): the mean value of 79 studies at pelagic sites was 103 mg CO 2 m −2 h −1 .If we take only the Chinese data from the database (n = 10), we have 72 mg CO 2 m −2 h −1 as the flux from the pelagic zone.In the present (littoral) data set, the total respiration rates are generally higher, averaging 346 mg CO 2 m −2 h −1 and of course much higher in the non-flooded area where vegetation flourishes.We conclude that the littoral zone is more active in terms of CO 2 exchange than the open water when compared on a per area basis, and should be taken into account when evaluating the role of water bodies in the carbon cycle, particularly in those reservoirs with a large ratio of littoral: open-water.

Influences of Environment
The CO 2 emissions differed significantly among the water levels, the months and the different times of day (Figure 2 and Table 2).Biomass and temperature were two important factors determining the spatiotemporal variations [33,34].For the flux differences among water levels, biomass showed the best fit (r 2 of linear regression was up to 0.99).Comparison between fluxes with and without aboveground vegetation of the 15 sites showed that shoot respiration was responsible for 27-75% of the total CO 2 emission.After removal of the aboveground biomass, the obvious differences among sampling sites under the same water condition vanished (Figure 3).For temporal variation, temperature showed a good fit.The correlations of temperature with fluxes during the different times of day were better than with that of different months.Fluxes of different months were also influenced by other factors, such as the biomass and soil water condition, which could alter the response of respiration to temperature [35].However, those factors were relatively stable in one day, showing a clearer temperature effect at a daily scale.
Besides biomass and temperature, water condition also influenced the CO 2 exchange.That was reflected in the significant difference between the emissions from flooded sites and non-flooded sites.On the one hand, flooding could inhibit autotrophic respiration by killing flood-intolerant vegetation through lack of oxygen and accumulation of toxic compounds [36,37].On the other hand, flooding could slow down heterotrophic respiration through limiting aeration [38].Aerobic respiration results in much more CO 2 production than anaerobic respiration: McNicol and Silver's study showed a 50% decrease of soil respiration caused by limiting aeration [39].
Additionally, change of the water condition could also influence C accumulation as biomass.The seasonally-flooded site was the only area which turned from carbon source to carbon sink in the autumn (Figure 4).The decline was caused by the increased accumulation of biomass.The important role that ephemeral vegetation plays in C fixation has been reported in the marginal area of a river [40].After water recession, new plants sprouted and grew on the exposed soil.When considering CH 4 flux separately, flooding boosted the emission [18], but when CO 2 exchange was taken into consideration, flooding played a positive role in C gain/loss through addition of biomass (i.e., flooding-induced C gain as biomass was greater than C loss as CH 4 in the present study, Figure 4), which is restricted by flooding time and plant phenology.If water recession occurred in the non-growing season, the process of water level fluctuation would merely increase the CH 4 emission but not the uptake of CO 2 .Plants growing in the littoral zone are generally herbaceous.Although extra biomass accumulated after water recession, how much of it can be fixed in the soil after death highly depends on content of recalcitrant compounds, especially lignin, and the weather and hydrology of the following months [41].

Comparison with Other Greenhouse Gases and the Source/Sink Question
The contribution of CH 4 and N 2 O to the balance of greenhouse gases flux was important (CH 4 was most important in the present study, Figure S2), especially for the two permanently flooded sites, i.e., DW and SW.When considering only CO 2 , the sites DW and SW were greenhouse gas sinks, absorbing 9.9 g C m −2 and 60.9 g C m −2 , respectively, in the growing season.When CH 4 and N 2 O are included in the calculation, DW and SW are revealed as greenhouse gas sources, losing the equivalent of 123.3 g C m −2 and 5.1 g C m −2 , respectively.The importance of CH 4 and N 2 O emission in greenhouse gas balances have been reported in numerous previous studies, for example coppice, grassland and cropland [42][43][44].A simulation study on the greenhouse gases balance of North American terrestrial ecosystems showed that emissions of CH 4 and N 2 O from wetland overturned the climate cooling effect induced by the CO 2 sink (2.2 times) and offset, by 0.3, 0.75 and 0.52 times, the sink effect of the forest, grassland and cropland respectively [45].Our study agrees with the ideas that all three gases should be accounted in global change research and non-CO 2 emissions could well be the dominant contributors in some wetland ecosystems.
For the growing season, the CO 2 , CH 4 and N 2 O exchange of the four different environments DW, SW, SF and NF was 123.3, 5.1, −70.0 and −109.1 g C m −2 , respectively (positive value indicates loss, negative value indicates gain).Our observed balances were more or less different from results of the previous studies.For example, the budget of a Danish riparian wetland was −450 to 40 g C m −2 (over 8 months) [46].A bog undergoing restoration in Canada absorbed 6 ± 28 g C m −2 year −1 [47].A peatland in Estonia lost 103 to148 g C m −2 year −1 [48].The comparison with the three temperate wetlands indicated that our estimation was within a reasonable range, although three reference ratios were taken to obtain the balance calculation, i.e., root to shoot ratio, carbon content of biomass and the ratio of heterotrophic respiration to soil apparent respiration.
Usage of an estimated reference values introduces uncertainties to varying degrees.The belowground biomass of 9 sites (4 replicates for each site, the 3 sites of DW were abandoned because of practical manipulation difficulties) was investigated in November 2011.To decrease disturbance to the sampling area, root biomass was measured only once.Root to shoot ratio ranged from 0.13 to 0.93, and the average was 0.41.Ratios of root to shoot are expected to change spatially and temporally along environment gradients, even for a given species, such as the soil water content and the available organic matter.The calculation of the C balance is sensitive to this ratio: if we take the extremes (0.13 and 0.93), the balance becomes 22.9 g C m −2 and −79.0 g C m −2 .
The carbon content of vegetation is relatively constant across a wide variety of species.Reference ratios are sometimes provided to enable simple carbon cycle calculations.The IPCC (Intergovernmental Panel on Climate Change) recommends 0.5 for wood and 0.4 for herbs [25].The FAO (Food and Agriculture Organization of the United Nations) recommend 0.475 for wood [49].In the present study, we assumed 0.4 as the vegetation is predominantly herbaceous.A study of the carbon content of 15 wetland plants of which 8 species are in the same family as the plants growing in our sampling area gives an impression of the likely variation: the carbon content was from 0.35 to 0.54 [50].If we replaced 0.4 by 0.35 and 0.54, the estimated average balance will be 9.7 and −75.6 g C m −2 .
Following a study on methods of NEE evaluation based on the chamber technique [22], the ratio of heterotrophic respiration to soil apparent respiration in the present research was assumed to be 0.55.In reality, this ratio may be expected to vary across time and space.Tang et al. reported a ratio of 0.56 during the growing season and 0.16 during the non-growing season with an annual average of 0.44 [51].Hanson et al. summarised 50 studies working with these ratios and found a wide range from below 0.1 to above 0.9 [52].If, in extremis, we were to take 0.1 and 0.9 for the present assessment, the balance would be −96.0g C m −2 and 52.0 g C m −2 .There may be other sources of error, for example the linear models (Table 3) used in evaluating the balance were themselves rather weak, especially the N 2 O model.However, the N 2 O emission (6.8 µg m −2 h −1 on average, see [17]) was small compared to that of CO 2 and CH 4 (1.3 mg m −2 h −1 on average, see [18]), and so we would not expect any big influence on the overall balance from the lack-of-fit of the N 2 O model.Finally, for the flooded areas, the flux observed after removal of shoots might include respiration of the phytoplankton.This fraction would have introduced uncertainties in the budget assessment since the ratio of heterotrophic respiration to soil apparent respiration was from work on non-flooded soil.
Notwithstanding the caveats mentioned above, our study in the littoral zone of the Miyun Reservoir supports the view that the margins of lakes may be hotspots of greenhouse gases emission.The study emphasizes how water level fluctuation may play an important role in influencing the emissions, e.g.increasing CH 4 emission through O 2 consumption by decomposing of plant tissue during flooding [18], increasing N 2 O emission by providing suitable moisture content during water recession [17,53], and decreasing CO 2 emission by regrowth of vegetation after water recession (this study).

Figure 1 .
Figure 1.Experimental design [17,18].WL: water level.The difference between high WL and low WL was caused by summer flooding.m H indicates meters in horizontal; m V indicates meters in vertical.The sites are grouped at different heights.DW: deep water site; SW: shallow water site; SF: seasonally flooded site; SFC: 'control site' for the seasonally flooded site; NF: non-flooded site.A, B and C denote samples from different vegetation types within each height band.For more details on water depth and other environmental parameters, seeTable 1, Figures 2 and S1.

Figure 1 .
Figure 1.Experimental design [17,18].WL: water level.The difference between high WL and low WL was caused by summer flooding.m H indicates meters in horizontal; m V indicates meters in vertical.The sites are grouped at different heights.DW: deep water site; SW: shallow water site; SF: seasonally flooded site; SFC: 'control site' for the seasonally flooded site; NF: non-flooded site.A, B and C denote samples from different vegetation types within each height band.For more details on water depth and other environmental parameters, seeTable 1, Figures 2 and S1.

Figure 3 .
Figure 3. Influence of aboveground plant on monthly and spatial variation of CO2 flux (mean ± SE).Subfigures show the CO2 flux of site DW (a,b), SW(c,d), SF and its 'control site' SFC (e,f), and NF (g,h) respectively, with and without aboveground plant.The left plots were flux with aboveground vegetation while the right plots were flux after aboveground vegetation being removed.Days between dashed vertical lines were flooding period when seasonal flooded area was under water.

Figure 3 .
Figure 3. Influence of aboveground plant on monthly and spatial variation of CO 2 flux (mean ± SE).Subfigures show the CO 2 flux of site DW (a,b), SW(c,d), SF and its 'control site' SFC (e,f), and NF (g,h) respectively, with and without aboveground plant.The left plots were flux with aboveground vegetation while the right plots were flux after aboveground vegetation being removed.Days between dashed vertical lines were flooding period when seasonal flooded area was under water.

Figure 4 .
Figure 4. Cumulative greenhouse gas fluxes, biomass and the balance of the growing season.Subfigures show the cumulative greenhouse gas fluxes, biomass and the balance of site DW (a), SW (b), SF (c) and NF (d) respectively.Fluxes of CH4 and N2O were converted by multiplying by Global Warming Potential.Balance was the difference in terms of C-equivalent between greenhouse gases emission and biomass.Balance excludes CH4 and N2O was the difference in terms of C-equivalent between CO2 emission and biomass.Positive values indicate C loss while negative values indicate C gain.

Figure 4 .
Figure 4. Cumulative greenhouse gas fluxes, biomass and the balance of the growing season.Subfigures show the cumulative greenhouse gas fluxes, biomass and the balance of site DW (a), SW (b), SF (c) and NF (d) respectively.Fluxes of CH 4 and N 2 O were converted by multiplying by Global Warming Potential.Balance was the difference in terms of C-equivalent between greenhouse gases emission and biomass.Balance excludes CH 4 and N 2 O was the difference in terms of C-equivalent between CO 2 emission and biomass.Positive values indicate C loss while negative values indicate C gain.

Table 2 .
ANOVA table to test the effects of water level, sampling month and time of day on CO 2 flux.