3.1.2 Changes in dominant phyla of prokaryotes and fungi
The 12 most abundant prokaryotic phyla (relative abundance higher than 1%) and 6 fungal phyla (> 0.1%) were selected to evaluate the effects of water table drawdown and drainage age on the most significant taxonomic shifts in the soil microbial community of degraded peatlands (Figure 2c and 2d, Table S1). We observed the relative abundances of all prokaryotic and fungal phyla were changed significantly with water table (p< 0.05 for all, Table S1). Remarkably, the relative abundances of Proteobacteria, Acidobacteri, Actinobacteria, Basidiomycota and Zygomycota significantly increased, whereas Chloroflexi, Bacteroidetes, Crenarchaeota and Ascomycota markedly decreased with the water table drawdown (Figure 2c and 2d). Compared with short-term drainage, Actinobacteria, Gemmatimonadetes and Nitrospirae showed more relative abundance in long-term drainage sites, while the higher relative abundances of Bacteroidetes, Crenarchaeota, Chytridiomycota and Glomeromycota were observed in the short-term drainage sites (p< 0.05 for all, Table S1).
3.1.3Soil microbiota as biomarkers
We subsequently analyzed whether we could discriminate samples with different water table based upon the microbiota composition using a random-forest machine learning analysis (Zhang et al., 2019). We performed a ten-fold cross-validation to evaluate the importance of indicator prokaryotic classes. We detected 14 classes as biomarker taxa when the cross-validation error curve was lowest (Figure 3a). Among theses, 8 classes had higher relative abundance in intermediate and high water table treatments (IWT and HWT), while 6 classes showed higher relative abundance in low-water table treatments (LWT; Figure 3b). This result demonstrates that soil microbial biomarker can be used as one of the important indexes to evaluate peatland degradation.
3.1.4 Functional annotation and distribution of prokaryotic community
FAPROTAX was further applied to evaluate the influence of water table on soil C and N processes (Figure 3c). A total of 69 functional pathways were obtained from 1688 OTUs. We found no significant difference in the relative abundances of functional pathway between the IWT and HWT, so we further analyzed the difference between the LWT and IWT and HWT treatments, and found that there were 42 functional pathways with significant differences (Figure 3c). The results showed that the low water table treatments had higher nitrate denitrification, nitrite denitrification, nitrate reduction and respiration, but had lower nitrite ammonification and nitrogen respiration. This imply that more NO3 were used for denitrification in LWT than IWT and HWT. In addition, we found the relative abundances of many carbon metabolism-related pathways, including methanogenesis, methanotrophy, fermentation, hydrocarbon degradation and etc., were significantly higher in IWT and HWT than in LWT.
GHG emission and itstemperature sensitivity
CO2 and N2O emission rates began low, peaked at about day 21, and kept a relative subsequently during the 35-day incubation. In addition, we found that 10 °C warming brought forward the peak of CO2 and N2O emissions by a week (Figure S3).
3.2.1GHG emission variation with water table and drainage age
The average GHG emission rate and its response to warming changed significantly with water table and drainage age (p < 0.01 for all, Table S2). At 8 °C and 18°C, the average rates of CO2, CH4 and N2O emissions from different water table treatments were significantly higher at short-term drainage sites than long-term drainage sites, respectively (Figure 4 and Table S2). At the same time, we also found that CO2 and CH4 emission rates were all significantly higher under IWT conditions than the other conditions at both two temperatures, while the N2O emission rates were significantly higher under LWT conditions (Figure 4 and Table S2). Similar to the results of FAPROTAX analysis, the GHG emission decreased with the increase of drainage age, and water table drawdown reduced the CO2 and CH4 emission rates, but increased N2O emission rates.
3.2.2 GHG emission variation with warming
Warming significantly increased the average rates of CO2 and N2O emissions of all treatments at short-term and long-term drainage sites, but slightly decreased the CH4 emission rates, even not significantly (Figure 4). We found the average CO2 emission rate at short-term drainage site was 36.34 ± 19.21 μg g-1 d-1 at 8 °C and 63.27 ± 31.18 μg g-1 d-1 at 18 °C, and significantly increased by 74.11% with ten degree warming, while the average CO2 emission rate at long-term drainage site was 24.25 ± 10.21 μg g-1 d-1 at 8 °C and 42.57 ± 10.75 μg g-1 d-1 at 18 °C, which significantly increased by 75.55% with 10 °C warming. In addition, we also found warming significantly increased average CO2 emission rate of LWT conditions by 91.14% from 19.46 ± 6.13 μg g-1d-1 of 8 °C to 37.78 ± 8.66 μg g-1d-1 of 18 °C, IWT conditions by 56.44% from 46.60 ± 13.20 μg g-1 d-1 of 8 °C to 72.90 ± 28.88 μg g-1 d-1 of 18 °C, and HWT conditions by 93.60% from 24.83 ± 3.56 μg g-1d-1 of 8 °C to 48.07 ± 5.64 μg g-1d-1 of 18 °C. The results of temperature sensitivity (Q10) of soil respiration also further proved that the sensitivity of IWT conditions to temperature were less than the other conditions (Figure S4).
Variation of DOC and TDN concentration during incubation
Compared with the soil before incubation, the variations of DOC and TDN concentrations after incubation showed different patterns (Figure 5). While TDN concentration increased, DOC concentration decreased with the water table drawdown at both short-term and long-term drainage sites during incubation (p < 0.001 for all; Figure 5 and Table S3). The DOC concentration of different water table treatments at short-term drainage site were significantly higher than long-term drainage site, respectively (p < 0.001; Table S3). Warming decreased the DOC concentration of all treatments (p < 0.05 for all) , but only significantly increased TDN concentration of LWT conditions (p < 0.05; Figure 5).
The influence factors ofGHG emissions
The average rates of CHG emissions during the incubation period was expressed in CO2, CH4 and N2O. CO2 and CH4 were positively correlated with DOC concentration under 8 °C and 18 °C (r = 0.555, r = 0.483 for CO2; r = 0.725, r = 0.555 for CH4; p < 0.05 for all), and were negatively correlated with TDN (r = -0.495, r = -0.505 for CO2; r =-0.449 for CH4 at 8 °C; p< 0.05 for all; Figure 6a, 6c and 6b,6d). No linear relationship was found among N2O and DOC, but N2O were correlated positively with TDN under 8 °C (r = 0.459, p < 0.05; Figure 6e and 6f).
Spearman’s correlation analysis were used to further evaluate microbial influences on the GHG emissions, DOC and TDN concentrations (Figure 7). An interesting result was found that the relationship between the dominant microbial phyla (11 most abundant prokaryotic phyla and 5 fungal phyla) and carbon and nitrogen mineralization showed opposite trends (Figure 7). Firmicutes, Chloroflexi, Chlorobi, Bacteroidetes and Crenarchaeota of prokaryotic phyla and Ascomycota, Chytridiomycota and Glomeromycota of fungal phyla showed a positive relationship with carbon mineralization (including CO2, CH4 and DOC), but a negative relationship with nitrogen mineralization (including N2O and TDN). Proteobacteria, Acidobacteria, Planctomycetes and Gemmatimonadetes of prokaryotic phyla and Basidiomycota, Zygomycota of fungal phyla showed a negative relationship with carbon mineralization (including CO2, CH4 and DOC), but a positive relationship with nitrogen mineralization (including N2O and TDN).
Finally, SEM was applied to access the direct and indirect effects of water table, drainage age, soil properties and soil prokaryotic and fungal communities on GHG emissions (Figure 8). Drainage can directly influenced CO2 and N2O emissions, with little or no directly effect on CH4 (Figure 8). Water table showed a directly negative influence on N2O, with little or no directly effect on CO2 and CH4. Drainage age were directly negative related to CO2, with little or no directly effect on CH4 and N2O. Drainage can indirectly affected GHG emissions by directly influenced plant biomass, TC, TN, DON, SWC, NH4+-N, microbial C:N and soil prokaryotic and fungal communities (Figure 8). Prokaryotic community showed a directly positive associated with CO2, CH4 and N2O emissions (Figure 8). SEM suggested that prokaryotic community exhibited a larger impact on GHG emissions than fungal community (Figure 8).
DISCUSSION
  1. Microbial characteristics varied in relation toduration ofdrainage and water table drawdown
The response of microbes to drainage is likely to depend on peatland type, the alteration of water table fluction, and the extent of spatiotemporal variation (Andersen, Chapman, & Artz, 2013; Krista Jaatinen et al., 2008; Minick et al., 2019; Peltoniemi, Fritze, & Laiho, 2009). Our study found that both water table and drainage age significantly affected the microbial community structure and compositions (Figure 2), which was consistent with many studies (Jaatinen et al., 2007; Urbanova & Barta, 2016). Drainage can directly lead to water table drawdown in peatlands, resulting in persistent aerobic conditions (Holden, Chapman, Lane, & Brookes, 2006), while with long-term drainage may further change the vegetation composition, thus affecting microbial composition and function (Miller, Benscoter, & Turetsky, 2015; Urbanova & Barta, 2016). Our results showed that water table drawdown significantly increased the relative abundances of Proteobacteria, Acidobacteria, Actinobacteria and Basidiomycota (Figure 2c and 2d). In oxic conditions, Members of Proteobacteria (Di Lonardo, De Boer, Klein Gunnewiek, Hannula, & Van der Wal, 2017), Acidobacteria(Dedysh, 2011), Actinobacteria (Goodfellow & Williams, 1983) and Basidiomycota (Ludley & Robinson, 2008) are considered to be the principal decomposers of soil organic matter, that can degrade recalcitrant organic materials (e.g. lignin, cellulose and humic materials) (Chen et al., 2018; Pankratov, Dedysh, & Zavarzin, 2006). So these microbes play an important role in GHG emissions, and this results consistent with previous research (Schimel & Gulledge, 1998). In addition, we also found long-term drainage significantly increased the abundance Actinobacteria of compared to short-term drainage. Relevant studies have found that Actinobacteria are more sensitive to long-term water table drawdown than short-term water table drawdown(Jaatinen et al., 2008; Peltoniemi et al., 2009).
Microbes are ubiquitous and mediate the macroscopic characteristics of the ecosystems (Thompson, Johansen, Dunbar, & Munsky, 2019). Because of the complexity of microbial communities make it necessary to explore functional relationships between specific microbes and ecosystem characteristics. Here, we have used machine learning techniques to find that specific microbial taxa could be used as one of the important indexes to evaluate peatland degradation. The relative abundance of 4 classes of the phylum Proteobacteria were significantly higher in the LWT compared to IWT and HWT (Figure 3b). As the members of Proteobacteria are involved in bacteriochlorophyll-dependent photosynthesis, they also considered to be the major decomposers of soil organic matter, and showed important role in nutritionally limited or arid environments (Ren et al., 2018). In addition, as an important member of Deltaproteobacteria, methanotroph can effectively oxidize methane in nature (Hanson & Hanson, 1996). In consistent with other studies (Cao et al., 2018; Siljanen, Saari, Bodrossy, & Martikainen, 2012; Zhong et al., 2020), we also found the relative abundance of Deltaproteobacteria increased with water table drawdown.
Influence of duration of drainage and water table on GHG emissions
The emissions of CO2, CH4 and N2O were significantly different between short-term drainage sites and long-term drainage sites (Figure 4 and Table S2). The long-term drainage peatlands are less suitable for microbial use compared with short-term drainage peatlands because of the poor quality of organic matter and low decomposability due to drainage over many decades of grazing history (Andersen et al., 2013; Leifeld, Steffens, & Galego-Sala, 2012; Urbanova & Barta, 2016). So far, few studies have examined the impact of duration of drainage on GHG emissions of peatland. Most researches have focused on the effect of drainage on GHG emissions in disturbed peatland compared to natural peatland (Cao et al., 2018; Maljanen, Hytönen, & Martikainen, 2001; Nieveen, Campbell, Schipper, & Blair, 2005; Zhou, Cui, Wang, & Li, 2017), or the effect of the water table drawdown on GHG emissions due to drainage (Chen, Borken, Stange, & Matzner, 2012; Hou et al., 2013; Laiho, Silvan, Cárcamo, & Vasander, 2001; Saurich et al., 2019; Wang, Siciliano, Helgason, & Bedard-Haughn, 2017). Kang et al. (2018) showed that N2O flux was found to be inconsistent under drought conditions, which may be caused by SOC variations by drought. This is in accordance with our results, We found N2O emission in long-term drained peatland was lower than that in short-term drained peatland (Figure 4). Huang, Zou, Zheng, Wang, and Xu (2004) found long-term drought conditions may induced the SOC to be recalcitrant, indicating a reduced supply of organic carbon for microbial activity. Therefore, nitrification and denitrification processes were probably inhibited to produce less N2O under the long-term drought conditions.
Many researches have found that water table drawdown caused by drainage make peatlands into sources for CO2 and N2O, whereas CH4emission is known to decrease(Cao et al., 2018; Saurich et al., 2019; Tiemeyer et al., 2016). In contrast, Knorr, Glaser, and Blodau (2008) and Muhr, Höhle, Otieno, and Borken (2011) found the emissions of CO2 from minerotrophic fen peatland had no change after water table drawdown. In accordance with other previous researches (Hou et al., 2013; Maljanen et al., 2001; Regina, Silvola, & Martikainen, 1999), we also found the N2O emission increased with the water table drawdown. The nitrogen cycle of peatland soil is very sensitive to the fluctuation of water table (Pal, Stres, Danevčič, Leskovec, & Mandic-Mulec, 2010). It is well known that soil nitrate is reduced to N2, NO and N2O through microbial processes (Knowles, 1982). N2 is the end product under anaerobic conditions, while N2 generally replaced by N2O at higher oxygen levels (Weil & Brady, 2017). The results of FAPROTAX also further supported our result, which we found had higher nitrous oxide denitrification, nitrate denitrification, nitrite denitrification, nitrate reduction and respiration, but had lower nitrite ammonification and nitrogen respiration (Figure 3c). In addition, we also found that multiple carbon metabolism-related pathways were significantly higher in IWT and HWT than those of LWT, which was partly supports the results of the carbon emissions (Figure 3c and Figure 4). Freeman, Lock, and Reynolds (1992) and Kwon et al. (2013) pointed out that water table drawdown changes anaerobic surface peat into aerobic, increasing the decomposition rates, microbial activity and aerobic respiration, and then increasing CO2 emissions. However, our results showed that CO2 emissions first increased and then decreased along the water table gradient. Similarly, Laiho (2006) showed that short-term water level drawdown increased CO2 emissions, but the longterm changes of the water level drawdown caused by the drought decreased the CO2 emissions of peatland. Hou et al. (2013) also found that water depth of 5 cm below surface increased the CO2 emission of peatlands, but the continued water table drawdown was no significant influenced or even decreased the CO2 emission. Swails et al. (2018) further proved that degradation of soil organic matter quality and nutrients associated with drainage may decrease substrate driven CO2 production from peat decomposition. Therefore, water table drawdown increased the emissions of CO2, but this impact maybe offset by long-term drought.
Influence of temperature on GHG emissions
Our study results consistent with previous studies, which found warming increased CO2 (Laine, Makiranta, et al., 2019) and N2O emissions (Cui et al., 2018; Duval & Radu, 2018), but have no significant impact on CH4 emission (Johnson, Pypker, Hribljan, & Chimner, 2013; Pearson et al., 2015). As for the impact of temperature on CO2, it is now generally accepted that warming can significantly increase CO2emissions in both drained and undrained peatlands (Hopple et al., 2020; Laine, Makiranta, et al., 2019; Salm, Kimmel, Uri, & Mander, 2009). However, warming appears to have a complex impact on the emission of CH4 from peatlands (Yang et al., 2014). Turetsky et al. (2008) found a positive correlation between warming and CH4 emissions, some reported a negative relationship (Eriksson, Öquist, & Nilsson, 2010; Peltoniemi et al., 2016), and while some other studies have showed no effect (Laine, Mehtatalo, Tolvanen, Frolking, & Tuittila, 2019; Pearson et al., 2015). Recenly, many studies have found that warming effect on CH4 emissions varied with the fluction of water table (Laine, Makiranta, et al., 2019; Peltoniemi et al., 2016; Yang et al., 2014). Gill, Giasson, Yu, and Finzi (2017) and Munir and Strack (2014) reported that warming increased the CH4 emisssion in water-saturated conditions, while the opposite phenomenon has been found from drier hummocks. This is partly consistent with our research, we only found that warming increased CH4 emission of HWT conditionin short-term drained peatlands, although not significant. Peltoniemi et al. (2016) concluded that the effect of warming under different moisture conditions on the activity and community of microorganisms regulating the methane cycle are not directly.
Consisitent with other ecosystems, for example alpine swamp meadow (Chen et al., 2017), high arctic tundra (Gong, Wu, Vogt, & Le, 2019) and subarctic tundra (Voigt, Lamprecht, et al., 2017), we found that warming increased N2O emissions with different water table treatments in both short-term and long-term drained peatlands at seasonal frozen soil. Butterbach-Bahl et al. (2013) reported that denitrification and consequently N2O emissions have high temperature sensitivity. The emission of N2O is controlled by of mineral nitrogen availability and influenced by microbial nitrification and denitrification processes (Bouwman, 1990). N mineralization from SOM can be accelerated by warming and leaching of nitrate and ammonium may occur (Bai et al., 2013). Thus, increased soil temperature promoted N2O emissions via denitrification(Voigt, Marushchak, et al., 2017). In addition, we also found Q10 values for GHG emissions were all lowest in IWT conditions (Figure S4). This indicates that water table is also an important index affecting the temperature sensitivity of GHG.
The role of biogeochemical factors in GHG emissions
Many researches have showed that drainage affected the processes of the carbon (C) and nitrogen (N) mineralization of peatland, and thus changed its C and N sink and sources function (Chen et al., 2012; Chimner, Pypker, Hribljan, Moore, & Waddington, 2017; Laine, Makiranta, et al., 2019; Zhang et al., 2020). The fluction of water table can directly affect soil biogeochemical properties, including soil basic physical and chemical properties, soil substrate, microbial community and enzyme activity, which lead to the change of GHG emissions (de Vries et al., 2018; Mpamah, Taipale, Rissanen, Biasi, & Nykanen, 2017; Swails et al., 2018; Wen et al., 2019). The concentration of DOC and TDN are the balance between soil organic matter production and soil microbial consumption, and DOC and TDN will be further decomposed and discharged in the form of CHG (van den Berg, Shotbolt, & Ashmore, 2012). Therefore, DOC and TDN concentration in soil can reflect carbon and nitrogen loss (Boothroyd, Worrall, & Allott, 2015). The results of our study found DOC and TDN concentration were strongly correlated with CO2 and CH4 emissions, while no linear relationship was found between N2O and DOC, N2O was only correlated positively with TDN (Figure 6). This suggests that soil substrate availability is crucial in GHG emission from soil. We further analyzed the relationship between soil microorganisms and substrate availability and GHG emission. An interesting result was found that some microbial taxa were contribute significantly to carbon mineralization, and others contribute significantly to nitrogen mineralization (Figure 7). Soil microorganisms affect the mineralization of soil organic C and N by secreting extracellular enzymes (Chapman, Cadillo-Quiroz, Childers, Turetsky, & Waldrop, 2017; Schnecker et al., 2015). Abatenh, Gizaw, Tsegaye, and Genene (2018) reported that microbial processes have a central role in GHG emissions and specific functional microorganisms are responsible for the related biochemical processes, and possibly a rapid response to climate change. In addition, SEM was used to analyze the impact of environmental factors on the emission of CHG, and it was found that the prokaryotic microbial community had the greatest impact on GHG, which also further supported the important role of microorganisms in soil GHG emissions (Singh, Bardgett, Smith, & Reay, 2010).
CONCLUSION
The emissions of CO2, CH4 and N2O response to drainage is inconsistent. CO2 and CH4 emission rates first increased and then decreased along the water table drawdown gradient, while N2O increased along the water table gradient. The emissions of CO2, CH4 and N2O from different water table treatments were significantly higher at short-term drainage sites than long-term drainage sites. The results of FAPROTAX analysis also supported this perspective. In addition, warming significantly increased the average rates of CO2 and N2O emissions of all treatments at short-term and long-term drainage sites, but not significantly decreased the CH4 emission rates. The Q10 values of GHG emissions were lowest in IWT conditions compared to other conditions. Microbial community composition was the primary factor affecting GHG emissions from peatlands, especially prokaryotes. Collectively, Our results further reveal the mechanism of climate change and human activities on the emisssions of GHG in alpine peatland ecosystem, which can provide support for the sustainable management of alpine peatland in the future.