Large dams may be substantial contributors to greenhouse gas emissions. Nitrous oxide (N2O) is the third most important greenhouse gas but studies on N2O emission from reservoirs are limited. We measured N2O emissions and environmental factors including atmospheric pressure, wind speed, air and soil/sediment temperature, biomass, soil water content and organic matter, total nitrogen, NH4+-N and NO3−-N of soil, from the littoral zones of the Miyun Reservoir, near Beijing, China, in January, May, June, August, and October during 2009 and 2010. Using the static chamber method we investigated the seasonal and spatial variation, relating it to environmental factors. Spatial and temporal variations in N2O flux appeared to be influenced by several environmental factors, working singly or in conjunction, including soil water depth, soil nutrition, biomass, and wind speed. In winter and spring, high N2O emissions (up to 1.9 ± 0.6 mg N2O m−2 h−1) were recorded at both eulittoral and infralittoral zones, while the flux from the supralittoral zone was low during all the seasons (from −0.04 to 0.01 mg N2O m−2 h−1). This study suggests that the littoral zone is a substantial source of N2O. However, its spatiotemporal variation and environmental drivers are still not clear.
Nitrous oxide (N2O) is considered to be the third most important greenhouse gas (Hernandez & Mitsch 2006; Ravishankara et al. 2009) with the current atmospheric concentration of 324 ppb and a warming potential of approximately 298 times that of CO2 over a 100-year time span (IPCC 2013). Over a decadal time scale, the atmospheric concentration of N2O has been increasing by 0.2–0.3% each year, mainly due to human perturbation of the global nitrogen cycle (e.g., Chen et al. 2011). It is believed that global warming and the deposition of nitrogen are increasing the flux of N2O from boreal lakes, and that such processes accelerate the release of N2O through positive feedback mechanisms (Liikanen et al. 2002; Cantarel et al. 2011; McCrackin & Elser 2011).
Wetland ecosystems have been suggested as important sources of N2O emission; in particular, many authors have focused on rice paddy wetlands which are a major N2O source (Kreye et al. 2007; Li et al. 2009). Limited studies on N2O emission of natural wetlands show big spatial variations. Nitrous oxide emission from rivers appears to be much lower than estuaries, contributing for example only 0.5–12.5% of the total N2O emission from some English and Welsh rivers and estuaries (Dong et al. 2004). Open water areas of natural lakes do not contribute significantly to N2O emissions either (Mengis et al. 1997), but N2O emissions from their littoral areas maintain relatively high levels (Senga et al. 2001; Huttunen et al. 2003; Wang et al. 2006).
Reservoirs are an important type of artificial wetland. The global area of reservoirs was estimated as 251,000 km2 (Lehner & Döll 2004) but may be much higher now. In 2000, 46% of all large dams in the world were found in China (Dams 2000). Under global climate change mitigation strategies, hydropower has often been considered as a clean energy source compared to fossil fuel; however, greenhouse gas emissions from reservoirs may have been underestimated (Fearnside 2002).
Reservoirs are associated exclusively with human development (Stadmark & Leonardson 2005) and are intimately linked to water dynamics, which in turn influences the nitrogen cycle, as a key controlling factor in N2O emissions (Hefting et al. 2004; Teiter & Mander 2005; Hernandez & Mitsch 2006). There have been an increasing number of studies on greenhouse gases released from reservoirs (e.g., Barros et al. 2011; Wehrli 2011). Most of these studies, however, have quantified emissions of CO2 and CH4, with little focus on N2O.
Greenhouse gas emissions from reservoirs in the tropics have attracted much attention (Galy-Lacaux et al. 1999; Rosa et al. 2004; Guérin et al. 2008) and reservoirs in other climatic regions have also been studied (Diem et al. 2008, 2012; Deemer et al. 2011). Lowland reservoirs were found to be major sources of N2O, while sub-alpine and alpine reservoirs appeared to be in equilibrium with atmospheric concentrations (Diem et al. 2012). Rather little is known about N2O emission from reservoirs in temperate regions, with data especially lacking on seasonal and diurnal patterns. Nitrous oxide production is influenced by a number of factors, such as nitrate concentration, organic matter availability, and pH (Hefting et al. 2003; Wu et al. 2009; Baulch et al. 2011). Reservoirs can include a wide range of primary production and redox conditions, so that N2O flux studies in reservoirs, in relation to environmental conditions, may contribute to a better overall understanding of the N2O flux in aquatic systems as well as contributing to our knowledge of the total greenhouse gas emissions associated with construction and use of reservoirs.
The objectives of the present study are: (1) to understand how N2O emitted from the Miyun Reservoir in China varies spatially and temporally; and (2) to assess key environmental factors contributing to any spatiotemporal variation in emission patterns.
The research was conducted at Miyun Reservoir (40°29′N, 116°50′E) near Beijing, China during 2009–2010. The reservoir is a 188 km2 water body and is up to 60 m deep. It has a capacity of 4.4 billion m3 and is the most important water resource for Beijing. The reservoir was constructed in 1960 by clearing vegetation and building a 66 m dam. Inflows to the reservoir are mainly from the Chao and Bai Rivers. The river catchments are characterized by warm-temperate semi-humid monsoonal climatic conditions with an average annual precipitation of 669 mm, with 80% of precipitation falling between July and September. The annual average regional air temperature, lowest air temperature, and highest air temperature is, respectively, 10.6 °C, −18 °C and 38 °C. There are, on average, 176 days annually without frost (Xie & Wang 2006). The reservoir is covered by ice approximately from mid-November to the end of March (Gao 1989). The littoral area includes rocky parts without vegetation and other parts which are muddy and vegetated. Eighty percent of the littoral area is muddy (Guan et al. 2007). Most of the muddy littoral area is distributed at the northern part of the reservoir (Figure 1), and we chose to study this part of the reservoir because it had the more typical habitat.
The study area was divided into three zones (Figure 1) delimited by water level and vegetation. The three zones were termed supralittoral (rarely inundated), eulittoral (intermittently inundated), and infralittoral (emergent only at extremely low water level), respectively (Wang et al. 2006). During the sampling period, the water level declined around two meters continuously. Because the slope was gentle, the water level fluctuation affects a large area. Details of the water depths in each of these zones are shown in Figure 2. The vegetation of the supralittoral zone was dominated by Artemisia capillaries Thunb., Eleusine indica (L.) Gaertn, Roegneria kamoji Ohwi, the eulittoral zone by Medicago sativa L. and Humulus scandens (Lour.) Merr., and that of the infralittoral zone by Typha orientalis Presl., Echinochloa caudata Roshev, and Echinochloa crusgalli L. For more details on weather, biomass, soil water content (SWC), and soil C/N, see Figure 2 and Table 1.
Flux measurements of N2O
Nitrous oxide samples for flux measurements were collected during the period from June 2009 to May 2010 (five collections over a 1-year period). For each of the three zones sampled, three replicate plots were established (n = 3). To avoid disturbance to the soil, especially trampling pressure, and compaction of soil, a wooden platform was built. All samples were collected at the same time of day (09:00–09:30) in June, August, October 2009, January, and May 2010. To analyze diurnal variation, samples were collected at 3-hour intervals on 4 May, 11 August, and 25 October. Each sampling was completed in a 3-day period, with one zone sampled per day.
Monitoring of environmental parameters
While collecting N2O samples in the field, several environmental parameters were also measured. Atmospheric pressure was measured with a barometer (DYM3, Baoping, China), and wind speed with an anemometer (4101, Testo, Germany) at the top of the chamber during sampling. Soil or sediment temperature at 5 cm below the surface, soil or sediment surface temperature, and air temperature inside and outside the chamber were recorded using a portable digital thermometer (JM624, Jinming, China) during collection of the gas samples.
After gas sampling, above-ground biomass was determined from three replicate harvests of 50 cm × 50 cm in June, August, October, and May, after drying at 80 °C to constant weight.
Soil water content of plots without standing water was measured using a Soil Water Sensor (UNI1000, Shunlong, China). Water depth of each of the plots with standing water was measured with a meter ruler at the same time that the gas samples were collected.
Soil samples of each replicate of the three areas were collected at 0–10 cm depth in May 2010. Total nitrogen (TN) content and soil organic matter (SOM) were monitored by the semi-micro Kjeldahl procedure and the potassium-dichromate oxidation procedure (Semenov et al. 2010). NH4+-N and NO3−-N were monitored by the potassium chloride-indophenol blue colorimetric method and the calcium sulfate disulfonic acid method (Lu 1999; Chen et al. 2010).
Two-way analysis of variance (ANOVA) was used to analyze the effects of sampling month, sampling zone, and the interaction of these two factors on N2O emission and environmental factors, including SWC, biomass, wind speed, soil temperature at 5 cm below the surface, soil surface temperature, and temperatures inside and outside the chamber in each sample zone. A one-way ANOVA was used to compare the fluxes among different sites in different months, and a Duncan's test was conducted to examine the difference in N2O flux among the treatments. Spearman correlation analysis was applied to determine the effect of environmental parameters on the N2O emission rate.
The relationship between the SWC and the N2O flux of each zone was analyzed separately by linear regression analysis. The SWC was considered 100% when there was water above the soil surface.
The effect of any variable was considered statistically significant if P < 0.05. All the analyses above were performed using IBM SPSS Statistics (19.0, IBM, USA). Charts were made using SigmaPlot (11.0, SYSTAT, USA) and Microsoft Excel (2010, Microsoft, USA).
Monthly and spatial variation in N2O flux
Two-way ANOVA analysis showed that month, sampling zone, and the interaction between these two factors all significantly affected N2O flux (P < 0.05; Table 2). Nitrous oxide emissions from all sampling zones in June, August, and October 2009 were significantly lower than in January and May 2010 except in the supralittoral zone. There was no significant difference among the sample zones for the June to October period (P > 0.05), although we noted that the flux was negative (i.e., N2O was absorbed) at the supralittoral zone in October (Figure 3). Nitrous oxide flux from the infralittoral zone in May was 1.9 ± 0.6 mg m−2 h−1, which was significantly higher (P < 0.05) than that from the eulittoral zone (0.7 ± 0.05 mg m−2 h−1), and the supralittoral zone, which was lower than zero (−3.8 ± 1.2 × 10−2 mg m−2 h−1). The same pattern was found in January for the different sampling zones (Figure 3).
Diurnal variation in N2O flux
Daily variation in N2O flux varied by both zone and time of year (Figure 4). For the supralittoral zone, in August the flux peaked at 09:00, then declined and was maintained at a slightly lower level until the next morning (Figure 4(a)). In October, the N2O flux remained near zero at the same zone (Figure 4(b)). However, this pattern changed greatly in May. Emissions first increased and peaked at 03:00, followed by a decline until 15:00, increasing again to reach its peak at 18:00 (Figure 4(c)).
For the eulittoral zone, nitrous oxide flux increased consistently from 03:00 peaking at 21:00 in August, compared to a relatively stable high emission level in October (Figures 4(d) and 4(e)). However, in May, N2O flux increased from 00:00 to 06:00, then increased and remained stable at a high level until 12:00; then, suddenly dropping to the lowest level at 15:00 (Figure 4(f)).
For the infralittoral zone, nitrous oxide flux first reached a peak at 06:00, and decreased to a relatively stable condition in August, but in October it peaked later, at 09:00 before decreasing to a stable level (Figures 4(g) and 4(h)). In May the flux was significantly higher and peaked at 15:00 (Figure 4(i)).
Key factors affecting N2O emission
The sampling month, sampling zone, and interaction of both factors significantly affected N2O emission (Table 2). We found environmental factors, including SWC, biomass, soil temperature at 5 cm below soil surface, soil surface temperature, air temperature within and outside the chamber, and wind speed were significantly different among seasons and among littoral zones (Table 2). Especially, the diurnal variation of wind speed presented the same pattern as the N2O flux in May in the eulittoral and infralittoral zones (Figure 4). The environmental parameters are shown in Figure 2. The atmosphere pressure was nearly the same at all the plots and months (Figure 2(a)), and the seasonal wind speed was not regular (Figure 2(b)). The temperature was the same pattern in all cases (Figures 2(c)–2(f)). Biomass from the infralittoral zone in August was the highest among all sites and months (Figure 2(g)). The SWC and water level were different among different sites and months (Figures 2(h) and 2(i)). The soil parameters, including SOM, TN, NO3− and NH4+ content were the highest in infralittoral zone (Table 1).
The Spearman correlation analysis indicated negative relationships between N2O emission and temperature and biomass (Table 3). There was no significant correlation between SWC and flux in infralittoral zone (R2 = 0.13, P > 0.05, df = 4; Figure 5) when SWC was lower than 20%. Negative correlations appeared when SWC was higher than 60% both in the eulittoral zone (R2 = 0.62, P > 0.05, df = 4) and the infralittoral zone (R2 = 0.97, P < 0.01, df = 4), but it was not significant in the eulittoral zone.
The present study showed significant variation in the spatial and temporal N2O flux (Table 2). Most notably, the mean N2O emissions recorded in the eulittoral and infralittoral zones were very high in January and May 2010 (Figure 3) when the SWC was low and there was little biomass (Figure 2).
From other work, SWC caused by variation in the level of groundwater is considered to be the dominant environmental factor affecting N2O flux (Breuer et al. 2000; Hefting et al. 2004; Hernandez & Mitsch 2006). For example, N2O emissions slowly increased with the SWC and water depth of the littoral marshes at the Three Gorges Reservoir (Chen et al. 2010). In our work, the N2O emission pattern emerges when the data are presented on an axis of soil moisture (Figure 5). The different patterns of emission of the three littoral zones shown in this graph might be attributed to soil anoxic conditions, which is a strong function of soil water. Facultative anaerobic microbes will not begin denitrification until O2 levels are insufficient (Batson et al. 2012). SWC normally affected both denitrification and N2O production (Senga et al. 2002; Dhondt et al. 2004). Both low levels of N2O flux and SWC suggested that the extreme low emission rates from the supralittoral zone might be caused by high available O2 in the soil pore space. In our study, N2O flux showed negative correlations when it was higher than 60% (Figure 5). These results agreed with research showing that N2O emission and SWC presented a parabolic correlation; and when the SWC was high, the strongly reductive condition in the soil allowed the complete reduction to N2, resulting in the decrease of N2O emission (Ciarlo et al. 2007). We propose that soil moisture explains the difference in N2O emission between June 2009 and May 2010, and also the higher N2O emission of winter (January 2010) than summer (August 2009). However, different results were obtained in a previous field study. A study carried out at riparian northern hardwood forests showed that there was no regular pattern between N2O flux and soil moisture (0.67 to 2.8 g H2O g dry soil−1) and instead flux was strongly related to percent plant cover (Hopfensperger et al. 2009). Although many studies have pointed out the importance of SWC to N2O flux, further field and laboratory work is needed to explain these inconsistent results.
Soil nutrients were considered to be the most likely explanation for higher emission from the infralittoral zone than the eulittoral zone. Denitrification is the only process that permanently reduces NO3− to N2 or N2O. The NO3− ion is utilized as a terminal electron acceptor by facultative anaerobic bacteria during respiration in the absence of O2 (Batson et al. 2012). Soil nitrogen, carbon, and C/N ratio have been found to be significantly correlated with denitrification potentials (Wu et al. 2013). Variation of denitrification primarily depends on nitrate (Song et al. 2012) and carbon availability (Song et al. 2011). Work at the Meiliang Bay of Taihu Lake, China, suggested hyper-eutrophication contributed to N2O emission (Wang et al. 2006). Similar results were found in a series of studies at the Yangtze River Delta, Huangpu River, and Chong-ming islands in the Yangtze River estuary (Hefting et al. 2003; Tomaszek & Czerwieniec 2003; Dong et al. 2004; Wang et al. 2009; Wu et al. 2009). Basing on such findings, we speculated the higher emission from the infralittoral zone than the eulittoral zone, at least in May, was attributed partly to the 2.1–5.5 times higher SOM and TN (Table 1). Moreover, spring rains may activate microbes (Breuer et al. 2000) or simply displace the greenhouse gas already stored in pore spaces (known as the ‘Birch effect’) (Unger et al. 2010).
The presence or absence of vegetation may affect emission rates in several ways. In the present study, low biomass was observed during high emission periods of January and May (Figure 2(g)). At these times, plants were either dormant or commencing growth, and so there is no strong competition with microbes for nitrogen use as there inevitably is during the period of growth (Zhang et al. 2009). Thus there would be more available nutrients left for N2O production, which might partly contribute to the higher emission of eulittoral and infralittoral zones in winter and spring. A similar observation was also shown in the study by Berglund & Berglund (2011) that N2O emission from cultivated peat soils with Lolium perenne peaked in springtime. Moreover, decomposition of residues of aquatic plants after the water level drops might supply substrates for N2O production, as at the Pantanal tropical wetland (Liengaard et al. 2013). Plant decomposition might also be the reason for the pattern of soil nutrients (Table 1). The biomass in the infralittoral zone was largest, where fixed carbon and nitrogen accelerated soil nutrient accumulation (Laurance et al. 1999; Sun et al. 2011). Although the biomass of the eulittoral zone was higher than that of the supralittoral zone, frequent water level fluctuation might aggravate the loss of nutrients through mineralization, nitrification, denitrification, and volatilization (Reddy et al. 1984), since the eulittoral zone is located at the edge of water and the higher ground.
Temporal variation in N2O emissions from soil can often be attributed to changes in temperature (Senga et al. 2002; Hefting et al. 2003; Dhondt et al. 2004). Some studies showed that temperature is positively correlated with N2O flux (Chen et al. 2010, 2011; Cantarel et al. 2011). However, negative correlation between N2O emission and soil and air temperature was observed in the present study (Table 3), which agreed with at least two other studies, one at a headwater stream and the other at a temperate farmland (Beaulieu et al. 2008; Cosentino et al. 2012). We propose this uncommon correlation occurs because the positive effect of temperature on decomposition is masked by even stronger effects from SWC and soil nutrients as discussed above.
Temperature is considered to be an important factor affecting N2O diurnal variation (Smith et al. 1998). However, we found wind speed could be the other factor influencing N2O flux (Figure 4). As shown in Figure 4, N2O flux totally or partly changed corresponding with wind speed especially in the period with high N2O flux. Wind affects air exchange velocity between soil/water and atmosphere (Wanninkhof & McGillis 1999; Reicosky et al. 2008), partly by increasing diffusion at the surface and a ‘bellows effect’ causing mass flow between pore spaces and the atmosphere through pressure fluctuations (Reicosky et al. 2008). Reicosky et al. (2008) showed that wind influences gas concentration at 30 cm depth in the soil which implies that a strong ventilation process is commonplace. Wind also is found to have an influence on soil respiration when using closed dynamic chambers (Bain et al. 2005). However, there is a lack of direct and specific research on the influence of wind on the flux measured by static chambers.
The average N2O emission rate from the littoral zone during this sampling period was calculated as 0.243 mg m−2 h−1 (varying from −0.1 to 1.9 mg m−2 h−1), which was similar to that found in previous studies in the littoral zone of lakes (Wang et al. 2006), but it is one to three times higher than the non-littoral areas of other reservoirs (Liu et al. 2011; Diem et al. 2012). That difference might be caused by the influence of water level and vegetation, as discussed above. This emphasizes the importance of the littoral zone to N2O emission from wetlands.
Previous studies provide data on other greenhouse gases from the Miyun Reservoir. Yang et al. (2011) found an average CH4 flux of 3.2 mg m−2 h−1 and a CO2 flux of 315 mg m−2 h−1. Considering the greenhouse effect of N2O is approximately 296 times that of CO2, and CH4 is approximately 23 times that of CO2 (IPCC 2001), nitrous oxide emission at the Miyun Reservoir was equivalent to 72 mg m−2 h−1 CO2, whereas the observed CH4 emission at the same site was equivalent to 74 mg m−2 h−1 CO2, showing the N2O emission at Miyun Reservoir represents a similar greenhouse effect as CH4 at the same zone.
Spatial and temporal variations in N2O may be influenced by several environmental factors, working singly or in conjunction, including soil water depth, soil nutrition, and biomass (Figure 6). This study provides some evidence of the underlying mechanisms which may help develop better management of the littoral zones at Miyun Reservoir in an effort to reduce greenhouse gas emission; and such management options may be transferred to other reservoir locations. For example, decreased frequency of water level fluctuation can shorten the duration of moderate humidity of soil. Of course, an integrated assessment is required to include other site-specific issues (e.g., water pollution and biodiversity) before implementation. This study recorded a substantial N2O flux which had a similar potential greenhouse effect as CH4 emissions at the same study site. We suggest the contribution of N2O to the ‘greenhouse effect’ might be as important as that of CH4, at least for the littoral area of a freshwater ecosystem. However, there remain high levels of uncertainty because of great spatiotemporal flux variation and environmental heterogeneity. For further works, more intense field observation and controlled experiments will help in developing a fully effective and workable management program to minimize greenhouse gas emissions from Miyun and other reservoirs.
This work was supported by the Fundamental Research Funds for the Central Universities (BLX2009-021 and YX2010-34); the State Forestry Administration of China under Grant 200804005; National Natural Science Foundation of China under Grant 31200313 and 31470475; and Ministry of Science and Technology of China under Grant 2013CB430405. We thank Dr Beth A Middleton, research ecologist in National Wetlands Research Center, US Geological Survey, for her suggestions on language. We offer our great appreciation to Professor John Grace from the University of Edinburgh, UK, for contributing ideas for the revision of this paper. We also thank Jingshi Wang, Kun Ma, Wei Wu, and Linlu Shi for their great support to this study. Meng Yang and Hongli Li contributed equally to this paper.