Effects of water level alteration on carbon cycling in peatlands

ABSTRACT Globally, peatlands play an important role in the carbon (C) cycle. High water level is a key factor in maintaining C storage in peatlands, but water levels are vulnerable to climate change and anthropogenic disturbance. This review examines literature related to the effects of water level alteration on C cycling in peatlands to summarize new ideas and uncertainties emerging in this field. Peatland ecosystems maintain their function by altering plant community structure to adapt to changing water levels. Regarding primary production, woody plants are more productive in unflooded, well-aerated conditions, while Sphagnum mosses are more productive in wetter conditions. The responses of sedges to water level alteration are species-specific. While peat decomposition is faster in unflooded, well aerated conditions, increased plant production may counteract the C loss induced by increased ecosystem respiration (ER) for a period of time. In contrast, rising water table maintains anaerobic conditions and enhances the role of the peatland as a C sink. Nevertheless, changes in DOC flux during water level fluctuation is complicated and depends on the interactions of flooding with environment. Notably, vegetation also plays a role in C flux but particular species vary in their ability to sequester and transport C. Bog ecosystems have a greater resilience to water level alteration than fens, due to differences in biogeochemical responses to hydrology. The full understanding of the role of peatlands in global C cycling deserves much more study due to uncertainties of vegetation feedbacks, peat–water interactions, microbial mediation of vegetation, wildfire, and functional responses after hydrologic restoration.


Introduction
Peatlands have an enormous potential role in global carbon (C) cycling, which depends on their hydrologic status (Moomaw et al. 2018). As a net sink for atmospheric CO 2 , peatlands store approximately 650 Gt (10 9 t) C (Yu et al. 2010;Page, Rieley, and Banks 2011), representing 20%-30% of the Earth's total soil organic C (Yu et al. 2010(Yu et al. , 2011Scharlemann et al. 2014). Boreal peatlands have sequestered approximately 270 to 547 Gt of C since the last deglaciation (Turunen et al. 2002;Yu et al. 2010;Loisel et al. 2014). Long-term C accumulation in peatlands is mainly ascribed to a long-term positive imbalance between two counteracting processes, C fixation from primary production and C loss from decomposition under high water level regime (Clymo 1984;Ballantyne et al. 2014), shown as vegetation inputs and microbial activities in Figure 1.
Peatland is by definition a type of wetland, with partially decomposed plant material (peat), hydrophytic plant species, and saturated environment . Though the classification of peatlands is still disputed, fen and bog are the two most widely accepted types. Fens are usually minerotrophic, which means they are mainly fed by nutrient-rich surface runoff and groundwater. Bogs are ombrotrophic or fed by precipitation and often isolated from the surrounding watershed. Another key difference of peatlands from other wetland types is that the vegetation of bogs and fens is generally comprised of trees, shrubs, sedges, grasses, and Sphagnum or brown mosses (Bridgham et al. 1996;Talbot 2009).
Peatlands are important in global C discussions with boreal, tropical, and temperate peatlands accounting for ~83.3%, 12.7%, and 4%, respectively, of global peatland area (Leifeld and Menichetti 2018). Most boreal and temperate peatlands are located in Europe, North America, and Russia under conditions of high precipitation and low temperature. Peat formation can also occur in tropical regions especially in climates with high precipitation and temperature (Page et al. 1999) including in Southeast Asia, Africa, the Caribbean, Central, and South America (Page, Rieley, and Banks 2011). Most tropical peatlands occur at low altitudes, where rainforest swamps have formed a thick mass of partially decomposed organic materials (Anderson 1983). Other tropical peatlands are found in upland or mountainous areas, some dominated by forest, and/or Sphagnum mosses (Page, Rieley, and Banks 2011).
The climates of boreal peatlands are predicted to become warmer (IPCC 2013), while studies also predict that annual precipitation will likely increase following warming (Arzhanov, Eliseev, and Mokhov 2012;Monier et al. 2013), especially during the winter (IPCC 2013;Screen 2013;Wu and Roulet 2014). This prediction results in complicated climatic patterns when simultaneously considering increased evapotranspiration, warming, and precipitation (Monier et al. 2013). The changes in summer precipitation can be regionally variable (Screen 2013), with both drought and flood occurring more frequently (Pal, Giorgi, and Bi 2004;Lehtonen, Ruosteenoja, and Jylha 2014;Wu and Roulet 2014). For temperate peatlands, larger but less frequent precipitation is projected (Sillmann et al. 2013;Wang et al. 2014a), while climate models also project that drought frequency and intensity are likely to increase, producing more extreme environments than in the past (IPCC 2013). Peatlands in Southeast Asia and North America are predicted to have both less precipitation and increased evaporation, leading to a drier environment (Li et al. 2007;Dai 2013). However, a study of flooding in the midwestern US reported an increase in the frequency of heavy rainfall days and stream flooding in the past 50 years, which may suggest that this region will be more flooded in the near-term future (Mallakpour and Villarini 2015). Because of the uncertainties of future hydrologic regimes in global peatlands, this paper considers the effects of both water table decline and rise.
Undisturbed peatlands serve as net sinks of atmospheric carbon dioxide (CO 2 ) and important sources of methane (CH 4 ), accounting for approximately 10% of global CH 4 emissions (Roulet 2000;Bridgham et al. 2013;Abdalla et al. 2016;Moomaw et al. 2018), even though emissions are seasonally and spatially variable (Roulet et al. 2007;Griffiths et al. 2017). Consequently, the peatland C cycle is an important part in the global C budget, because this cycle influences the dynamics of global warming (Artigas et al. 2015;Fleischer et al. 2016). Also see feedbacks to climate change in Figure  1. Since C balance in peatlands is sensitive to both water availability and temperature shifts (Turetsky et al. 2008;Kang et al. 2016), C cycle can be severely affected by climate change due to changes in temperature and precipitation patterns (Kettles and Tarnocai 1999;Li et al. 2007; Gallego-Sala and Prentice 2013; Walker et al. 2016).
In addition to climate change, land-use and landcover change (LUCC) related to anthropogenic disturbance can cause shifts in water levels and peatland functions (Ojanen, Minkkinen, and Penttila 2013;Renou-Wilson et al. 2014). Globally, approximately 10% of the world's peatlands have been drained or mined Joosten 2009;Leifeld and Menichetti 2018). In Europe, more than half of peatlands have been drained for agriculture or forestry (Dixon et al. 2014). Since the 1970s, approximately 6.0 million ha of peatlands have been drained on Finland (Menberu et al. 2018). Similarly, extensive peatland areas in the tropics have been drained with ditches, which results in increased emissions of CO 2 and extensive loss of peat due to oxidation (Page et al. 2009;Hooijer et al. 2012;Swails et al. 2019a). Drained peatlands in Indonesia emit the largest amount of CO 2 , which is equal to 8% of global fossil fuel emissions (Hooijer et al. 2006;Biancalani and Avagyan 2014). Drainage is estimated to cause a cumulative loss of 80.8 Gt C around the world (Leifeld and Menichetti 2018). Also, drainage for mining for agricultural and industrial usage lowers the water level of peatlands (WindMulder, Rochefort, and Vitt 1996; Price and Water level is the key factor in regulating peatland C cycling for its strong impact on vegetation and microbial activities. C balance in peatlands could affect climate change, which in turn influences peatland C storage via water level alteration. Whitehead 2001; Drollinger et al. 2020). Reservoir construction can elevate local water levels and flood peatlands (Kelly et al. 1997;St. Louis et al. 2000). The construction of roads or railways often creates embankments, which either may impound or dry peatlands (Harms et al. 1980;Patterson and Cooper 2007). Peatland restoration via ditch filling or dam construction can rewet peat and raise water level (Cui et al. 2017). Beaver dams also impound peatlands and are an important part of landscapes in boreal regions (Mitchell and Niering 1993;Moore et al. 2011).
The objective of this review is to explore the literature related to the effects of water level alteration by either natural or artificial means on C cycling in peatlands over short-term and long-term time frames. The review synthesizes new ideas and uncertainties emerging in this field. To develop this review, we gathered information from published studies of C cycling in both flooded and dried conditions including, laboratory, field, and modeling studies conducted in boreal to tropical ecozones. The main body of this paper was divided into three sections, representing responses of the three main components of C cycling in peatlands: vegetation (input), decomposition (output) and C storage (net effects).

Effects of water level alteration on carbon cycling processes
In peatlands, anoxia induced by high water levels defines vegetation composition and decomposition, thus controlling ecosystem function (Laine, Vasander, and Laiho 1995;Blodau, Basiliko, and Moore 2004;Strack, Waddington, and Tuittila 2004;Riutta, Laine, and Tuittila 2007;Strack et al. 2006a;Strakova et al. 2012). Temperature is also an important regulator of processes linked to photosynthesis, autotrophic respiration, and soil respiration (Arft et al. 1999;Medlyn et al. 2002;Poorter et al. 2012). However, the effects of temperature are controlled by alterations in water levels and associated oxygen levels, which makes water level a stronger driver of C cycling processes than temperature alone (Oechel et al. 1998;Davidson and Janssens 2006;Elmendorf et al. 2012;Munir et al. 2015;Buttler et al. 2015;Makiranta et al. 2018;Laine et al. 2019aLaine et al. , 2019b.

Effects of water level alteration on vegetation
In this section, we focused on the responses of the three major plant types of peatlands: Sphagnum mosses, graminoids (mainly sedges), and woody plants (shrubs and trees). Their key differences in ecophysiological traits lead to different regulatory impacts on ecosystem C dynamics.

Effects of water level alteration on vegetation composition
Water level is a predominant driver of peatland vegetation succession (Laine, Vasander, and Laiho 1995;Bunting and Warner 1998;Tuittila et al. 2000b;Charman 2002;Bubier, Moore, and Crosby 2006;Strack et al. 2006a;Talbot 2009;Laing et al. 2014;Malhotra et al. 2016), so that even a short-term decline in water level can drive species shifts (Maestre, Valladares, and Reynolds 2005;Garssen, Verhoeven, and Soons 2014). For example, Sphagnum mosses are key species in C accumulation in peatlands. Due to the lack of stomata and water-conducting tissues, Sphagnum depends on a constant water supply from precipitation and/or a high water table (Thompson and Waddington 2008;McCarter and Price 2014;Nijp et al. 2014). Wetter conditions favor the expansion of lawn/hollow species (e.g., Sphagnum angustifolium, Sphagnum rubellum), while drier conditions favor hummock species (e.g., Sphagnum capillifolium, Sphagnum fuscum) in North America (Andrus, Wagner, and Titus 1983;Maanavilja et al. 2015;Norby et al. 2019) and Europe (Robroek et al. 2007a;Kotiaho et al. 2013). Compared to lawn species, hummock-forming Sphagnum was more resistant to drought (Hayward, Clymo, and Fogg 1982;Hájek and Beckett 2008;Strack and Price 2009;Kuiper et al. 2014) with higher stem bulk density and more efficient water transport via capillary action (Grosvernier, Matthey, and Buttler 1997;Robroek et al. 2007b). Generally, water level drawdown increases moisture stress causing shrinkage of Sphagnum cover in peatlands (Weltzin et al. 2000(Weltzin et al. , 2003Strack et al. 2006a;Breeuwer et al. 2009;Potvin et al. 2015;Goud, Watt, and Moore 2018;Duval and Radu 2018;McPartland et al. 2019). Also, long-term drought can drive succession by suppressing Sphagnum and favoring the feather mosses of forested peatlands (Laine, Vasander, and Laiho 1995;Minkkinen et al. 1999;Rydin, Jeglum, and Bennett 2013;Kangas et al. 2014). The replacement of Sphagnum with vascular plants under water table drawdown has been reported in fens (Churchill et al. 2015) and bogs (Munir et al. 2015). However, Sphagnum mosses can lessen the effects of severe declines in water levels since they transport and store water by capillary action, which ameliorates local soil drying and maintains anoxia (Laine et al. 2011;Ketcheson and Price 2014). Rewetting peatlands benefits Sphagnum and helps to rebuild Sphagnum vegetation in restoration projects (Tuittila et al. 2000b), but overly wet environment can also suppress Sphagnum and lead to replacement by hydrophyte species that are better adapted to submergence (Granath, Strengbom, and Rydin 2010;Maanavilja et al. 2015;Chimner et al. 2016).
It is still uncertain how sedges respond to water level alteration. Some species of sedges are able to maintain abundance in high moisture (Bakker et al. 2007;McPartland et al. 2019), although dry conditions were also reported to advantage other sedge species (Strack et al. 2006a;Strack, Waller, and Waddington 2006b;Dieleman et al. 2015). Increased abundance of sedges has been observed under raised water table in restoration Tuittila et al. 2000b) and manipulation experiments (Weltzin et al. 2003;Churchill et al. 2015;Olefeldt et al. 2017), while mesocosm experiments found gradual replacements of Sphagnum mosses by sedges under drying (Weltzin et al. 2000;Breeuwer et al. 2009;Dieleman et al. 2015). However, water level drawdown also reportedly decreased sedge cover in both fens and bogs (Laine, Vasander, and Laiho 1995;Weltzin et al. 2003;Breeuwer et al. 2009;Potvin et al. 2015;Duval and Radu 2018).
Since woody species are generally shallow-rooted (Wallén 1986) and lacking in well-developed aerenchyma (Kozlowski 1997), the woody plant typically shows dominance in drawdown, well-aerated environments. Meanwhile, increased oxygen availability to microbes promotes peat mineralization and consequently elevates nutrient levels of the environment, which benefits nutrient uptake in woody plants more than Sphagnum species (Laiho 2006;Strakova et al. 2010;Bu et al. 2011;Strakova et al. 2012;Bragazza et al. 2013;Macrae et al. 2013). Shrubs increase in abundance as water tables decline (Weltzin et al. 2000(Weltzin et al. , 2003Breeuwer et al. 2009;Murphy, Laiho, and Moore 2009;Bragazza et al. 2013;Munir et al. 2015;Potvin et al. 2015;Duval and Radu 2018;Ratcliffe et al. 2019). Even a slight decline in water level may lead to the spread of certain shrub species such as Betula nana (Laine, Vasander, and Laiho 1995;Makiranta et al. 2018). In some natural peatlands, woody species are more abundant in raised hummocks that create well-aerated conditions for their growth (Chimner and Hart 1996). Shrubs can colonize peatlands, which can eventually facilitate a shift in vegetation progression from moss to shrub-dominated systems (Laine, Vasander, and Laiho 1995;Weltzin et al. 2000;Holmgren et al. 2015). Long-term drainage further leads a shift to tree dominance in bogs and fens (Laine, Vasander, and Laiho 1995;Minkkinen et al. 1999;Murphy, Laiho, and Moore 2009;Strakova et al. 2012;Munir et al. 2015;Chimner et al. 2016). The exact turning point in water level that promotes forest colonization during longterm drainage and climate change is yet unknown. After forest colonizes, the increasing dominance of species favored by well-aerated conditions will change the ecosystem function of the peatland (Laiho et al. 2003;Ratcliffe et al. 2017).
Sedge production has no single response pattern to water level alteration. Previous studies have reported no difference (Potvin et al. 2015;Chimner et al. 2016), decrease (Schreader et al. 1998;Weltzin et al. 2000;Bubier et al. 2003;Chimner and Cooper 2003a;Bragazza 2006;Schimelpfenig, Cooper, and Chimner 2014;Talbot et al. 2014;Churchill et al. 2015) and increase (Strack, Waller, and Waddington 2006b) in sedge production in lower water levels, depending on species. Wetter conditions enhanced photosynthesis of Eriophorum vaginatum, Eriophorum angustifolium and Rhynchospora alba in field studies Laine et al. 2009;Lazcano et al. 2020), whereas the biomass of Bolboschoenus planiculmis decreased in deeper water in both field and greenhouse experiments (Liu et al. 2018). Sedge species display various responses to water level alteration, because they have diverse hydrological and chemical niches (Chimner et al. 2016). The responses of sedge production to water level alteration not only depend on water availability, but also on changed nutrient conditions as hydrology changes (Riutta, Laine, and Tuittila 2007).
There is so far no general agreement on the response of total plant production to water level alteration. Decreased plant production in peatlands has been reported if water levels decline, especially during the growing season (Bubier et al. 2003;Chimner and Cooper 2003a;Blodau, Basiliko, and Moore 2004;Chivers et al. 2009;Lund et al. 2012;Churchill et al. 2015;Potvin et al. 2015;Chimner et al. 2016;Olefeldt et al. 2017;Kang et al. 2018), which is related to increased water stress (Alm et al. 1999;Griffis, Rouse, and Waddington 2000;Weltzin et al. 2000). Rewetting via hydrologic restoration can lead to an increase in production for increasing the production of dominant species that favors wetter conditions (Maanavilja et al. 2015;Karki et al. 2016), although drying can also increase plant production if the production of trees and shrubs increases (Sulman et al. 2009(Sulman et al. , 2012Munir et al. , 2015Järveoja et al. 2016;Kasimir et al. 2018;Ratcliffe et al. 2019). Notably, plants can change carbon allocation to respond to altered moisture conditions, with increased belowground primary production under drying and more aboveground primary production when water supply is sufficient, which thereby may cause no change in total primary production (Weltzin et al. 2000). Studies on gross photosynthesis show decreases under drought or peatland drainage and increases after rewetting Riutta, Laine, and Tuittila 2007;Aurela et al. 2009;Adkinson, Syed, and Flanagan 2011;Peichl et al. 2014;Humphreys et al. 2014), whereas lowered water levels with adequate moisture support increased gross photosynthesis and ER (Strack et al. 2006a;Strack and Waddington 2007;Cai, Flanagan, and Syed 2010;Laine et al. 2019a). NDVI (Normalized Difference Vegetation Index) is a typical proxy of vegetation greenness, which is positively correlated with peatland ecosystem productivity (Boelman et al. 2003(Boelman et al. , 2005. Decreased NDVI was observed with long-term reduction in the soil moisture of a rich fen, along with lower proportion of photosynthetic to non-photosynthetic leaf area (McPartland et al. 2019). Similarly, decreased vegetation LAI (Leaf Area Index) occurred in peatlands drained for forestry (Laine et al. 2016(Laine et al. , 2019b. After hydrologic restoration, LAI and vegetation composition may become more like undisturbed peatlands (Laine et al. 2019b).
The vegetation of bogs tends to be more resilient to water level alteration than fens, because bogs regularly undergo seasonal drought, while water input to fens is more continuous (Thormann, Bayley, and Szumigalski 1998). In the meantime, the nutrient-rich conditions of fens relative to bogs benefits species invasion as hydrology changes Eskelinen and Harrison 2014). As an example, obvious vegetation succession in fen communities occurs after both experimental drawdown and drainage for forestry, whereas bog communities do not change in similar circumstances (Minkkinen et al. 1999;Kokkonen et al. 2019).
Studies of the effects of water level alteration on vegetation reveal that the relationship is complex and depends on the nature of the hydrologic change, vegetation type, and nutrient levels (Chimner et al. 2016;Cao et al. 2017a). A diverse ecosystem with plant functional types adapted to new environmental conditions may maintain primary production after hydrological changes (Turetsky et al. 2012), suggesting that peatland ecosystems may be functionally resilient to future climates (Heijmans et al. 2008;Dise 2009;Wang, Richardson, and Ho 2015;Makiranta et al. 2018). Given the fact that most of the knowledge on vegetation response to hydrologic alteration has come from boreal and temperate peatlands (Mitsch et al. 2009), more studies on tropical peatlands would help our understanding of peatland responses and accordingly, help us to optimize vegetation management in peatlands under climate change.

Effects of water level alteration on organic matter decomposition
In peatlands, the decomposition of organic matter is controlled by oxygen (O 2 ) availability and diffusivity as related to hydrological regime, with much impact on the levels of emission of greenhouse gases from peatlands to the atmosphere (Elberling et al. 2011;Yvon-Durocher et al. 2014;Dickopp, Lengerer, and Kazda 2018). See aerobic and anaerobic decomposition in Figure 1. Drawdown alleviates oxygen limitation and increases soil aeration, which promotes the activities of soil organisms and microbial mineralization of organic substrates in peatlands (Moore and Dalva 1997;Laiho et al. 2001;Makiranta et al. 2009;Hribljan et al. 2014;Bragazza et al. 2016;Schmidt et al. 2016), as well as the activities of phenoloxidase and peroxidase (Fenner and Freeman 2011;Peacock et al. 2015;Dieleman et al. 2016a), hence leading to an increase in hydrolase activity and decomposition rate (Moore and Knowles 1989;Silvola et al. 1996;Nykanen et al. 1998;Minkkinen et al. 1999;Freeman, Ostle, and Kang 2001;Jaatinen et al. 2008;Peacock et al. 2015;Wu et al. 2017). For example, phenolic-degrading bacteria are more diverse and have higher phenoloxidase activity under short-term drought, which further increases CO 2 flux (Fenner, Freeman, and Reynolds 2005). Saprotrophic fungi play a dominant role in aerobic decomposition of peatlands (Latter, Cragg, and Heal 1967;Williams and Crawford 1983), which show greater richness under drier conditions, and they also exhibit high tolerance to drought (Yuste et al. 2011;de Vries et al. 2012;Barnard, Osborne, and Firestone 2013;Bragazza et al. 2015;Asemaninejad, Thorn, and Lindo 2017;Jassey et al. 2018). Meanwhile, drawdown modifies the structure of the microbial community and increases microbial turnover, thus enhancing enzymatic secretions, and subsequently microbial decomposition (Lamentowicz et al. 2013;Trap et al. 2016;Reczuga et al. 2018). Therefore, water levels in peatlands are a critical aspect of the dynamic between greenhouse gas emissions and decomposition rates (Moomaw et al. 2018).
Meanwhile, experiments have confirmed that water level decline has a stronger impact on CO 2 exchange than warming (Oechel et al. 1998;Chivers et al. 2009;Munir et al. 2015;Pearson et al. 2015;Laine et al. 2019a), and a significant increase in ER has been observed in peatlands when warming is combined with drawdown (Samson et al. 2018;Laine et al. 2019b). The effects of dry conditions following drawdown on decomposition can be neutral or negative if the environment (e.g., extreme drought or low temperature) reduces microbial activity (Updegraff et al. 2001;Allison and Treseder 2008;Lazcano et al. 2020), extracellular enzymes (McLatchey andReddy 1998;Toberman et al. 2008;Sinsabaugh 2010;Reczuga et al. 2018) or root growth (Palta and Nobel 1989;Rochette, Desjardins, and Pattey 1991;Atkin and Macherel 2009). In contrast, flooding may increase decomposition following the thaw of permafrost, which releases a large amount of labile C (Turetsky, Wieder, and Vitt 2002;Turetsky et al. 2007;Myers-Smith et al. 2007;Schuur et al. 2008;Kane et al. 2010).
Elevated CO 2 emission after water table alteration may further influence vegetation composition and production in peatlands (e.g., CO 2 fertilization), but these feedbacks are still poorly understood. Results of previous studies are conflicting and dependent on plant functional types. Vascular plants, especially sedges, show increased productivity and cover from CO 2 fertilization (Grulke et al. 1990;Fenner et al. 2007b;Dieleman et al. 2015), or alternatively no change , while there is a negative (Fenner et al. 2007b)  Sphagnum. The effects can be further complicated because of the interaction of CO 2 and warming (Long 1991;Luo et al. 2008;Dieleman et al. 2015). A better understanding of the relationship between the multiple drivers of primary production would be helpful in peatland management under climate change.

Effects of water level alteration on CH 4 emissions
Methanogenesis in peatlands is strongly controlled by anaerobic conditions under high water level regime (Conrad 2005;Rydin, Jeglum, and Bennett 2013). Methane is a greenhouse gas 45 times more potent than carbon dioxide in projections of sustained-flux global warming potential over the next 100 years (Neubauer and Megonigal 2015); however, the potential of methane in global warming is mitigated by its shorter half-life in the atmosphere (IPCC 2013). Methane is produced by methanogens during the final stage of anaerobic degradation of organic matter (Whalen 2005). CH 4 emission is thus sensitive to hydrologic change ) and is dominated by three pathways (1) diffusion through the acrotelm ), (2) ebullition (i.e., bubble release from saturated peat) (Rosenberry et al. 2003;Kellner, Price, and Waddington 2004), and (3) diffusive flux or internal pressurized gas flow through the aerenchyma of vascular species (Brix, Sorrell, and Orr 1992;Chanton et al. 1992;Whiting and Chanton 1996;Whalen 2005;Hornibrook 2013;van den Berg et al. 2016). Since the latter two processes make the greatest contribution to the CH 4 flux in peatlands (Whalen, Reeburgh, and Sandbeck 1990;Lai 2009), the changes in peat aeration, compaction and temperature as well as nutrient level and vegetation cover can substantially influence CH 4 emissions (Berger et al. 2018). Lowered water levels reduce CH 4 emissions by decreasing the abundance of methanogens and CH 4 production and increasing CH 4 oxidation levels, while raised water levels have a higher potential of anaerobic CH 4 production (Frenzel and Funk et al. 1994;Aerts and Ludwig 1997;Blodau, Basiliko, and Moore 2004;Strack and Waddington 2007;Turetsky et al. 2008;Chivers et al. 2009;Dinsmore et al. 2009;Laine et al. 2009;Pearson et al. 2015;Olefeldt et al. 2017;Kang et al. 2018;Yang et al. 2019;Wu et al. 2020). See also Table 2. However, water level decline may increase CH 4 emissions when it is associated with increased belowground root growth of aerenchymous species (Weltzin et al. 2000;Strack, Waller, and Waddington 2006b). In contrast, peatland restoration with rewetting leads to higher CH 4 emissions (Komulainen et al. 1998;Hendriks et al. 2007;Waddington and Day 2007;Worrall, Bell, Table 2). After hydrology is restored, the quantity of methanogens increases (Urbanová, Picek, and Bárta 2011), and CH 4 production is mainly controlled by nutrient supply (Couwenberg 2009;Lai 2009;Saarnio, Winiwarter, and Leitão 2009) and soil temperature (Schrier-Uijl et al. 2010).

Effects of water level alteration on DOC
Dissolved organic carbon (DOC) is exported through runoff and leaching from peatlands, and these processes contribute up to 25% of overall peatland C fluxes and 20% of terrestrial-derived DOC delivered to the oceans (Fenner et al. 2007a;Yu 2012). The majority of DOC exported from peatlands will be subsequently mineralized and converted to CO 2 (Wit et al. 2015;Evans, Renou-Wilson, and Strack 2016;Jones et al. 2016). This kind of waterborne C flux represents a primary factor in peatlands switching from net C sinks to net C sources (Fraser, Roulet, and Moore 2001;Billett et al. 2004), which has been confirmed by flux studies in peatlands (Roulet et al. 2007;Nilsson et al. 2008;Dinsmore et al. 2010;Koehler, Figure 1. Lower water levels facilitate oxygen diffusion and microbial decomposition, which promotes DOC production, while elevated water levels transport newly produced DOC from peat to pore water (Chow et al. 2006;Clark et al. 2009;Fenner and Freeman 2011).
There is no clear pattern of DOC concentration response to water level change. Higher DOC concentrations were measured in lower water depths in comparison to the control and higher water level sites in fens, bogs, and high-altitude peatlands, related to increased vegetation biomass or peat decomposition Kane et al. 2010;Lou et al. 2014;Armstrong et al. 2015;Strack, Munir, and Khadka 2019;Lazcano et al. 2020). Elevated DOC concentrations were also observed in tropical peatlands recently drained for forestry (Yupi et al. 2016) and in boreal peatlands after long-term drainage (Frank et al. 2014;Menberu et al. 2017), while rewetting through restoration lowered DOC concentrations in temperate (Holl et al. 2009) and boreal peatlands (Turner, Worrall, and Burt 2013;Menberu et al. 2017), corresponding to reduced peat decomposition. However, laboratory and field experiments showed that drawdown decreased DOC solubility and increased microbial consumption on DOC, thereby causing relatively lower DOC concentrations in peat; at the same time, DOC concentrations increased with peat water saturation (Fenner, Freeman, and Reynolds 2005;Clark et al. 2009;Preston, Eimers, and Watmough 2011;Urbanová, Picek, and Bárta 2011;Dieleman et al. 2016b). Higher DOC concentrations are found in both intact (Kalbitz and Geyer 2002) and restored peatlands (Zak and Gelbrecht 2007;Wilson et al. 2011;Strack et al. 2015), but not in degraded peatlands. In hydrologically altered peatlands in Michigan, DOC concentrations were higher in both drier and wetter sites than in the sites with unaltered hydrology . A meta-analysis based on 110 independent studies found that there is no significant effect on DOC concentration after either the drainage or restoration of peatlands (Haddaway et al. 2014). Under water level alteration, peatland DOC concentrations are dependent on multiple factors (e.g., changes in peat properties, magnitude and duration of hydrologic disturbance, initial vegetation composition and subsequent changes, nutrient status) and their interactions (Kalbitz and Geyer 2002;Zak and Gelbrecht 2007;Armstrong et al. 2012;Strack et al. 2015;Mastný et al. 2018;Lazcano et al. 2020), which makes water level alone not a strong predictor of DOC status .
A general pattern of DOC quality under water level alteration is that drainage increases DOC aromaticity and decreases DOC biodegradability (Wickland, Neff, and Aiken 2007;Hribljan et al. 2014;Hansen et al. 2016;Jassey et al. 2018). Specifically, water table drawdown increases the occurrence and molecular weight of phenolics in DOC (Blodau and Siems 2012;Frank et al. 2014;Lou et al. 2014;Kane et al. 2019), while rising water table decreases aromaticity and makes DOC more labile (Holl et al. 2009;Hribljan et al. 2014;Dieleman et al. 2016b).
DOC export is controlled by runoff, with higher DOC export generally found after peatland drainage (Joensuu, Ahti, and Vuollekoski 2002;Evans, Renou-Wilson, and Strack 2016). DOC exports were higher after experimental water level decline in boreal, temperate, and high-altitude peatlands, as water flux and fluctuation also increased Lou et al. 2014). Tropical peatlands have higher DOC export after drainage for forestry or plantation, which account for more than 90% of total organic carbon fluxes in these systems (Moore et al. 2013;Yupi et al. 2016;Cook et al. 2018). DOC export from peatland drained for forestry is one of the most important forms of anthropogenic C loss in boreal and tropical regions (Turunen 2008;Moore et al. 2013;Yupi et al. 2016). In this situation, maintaining high and stable water levels is more important than simply controlling CO 2 emissions from peatland surface, especially because the effluxes of DOC in drained tropical peatlands are mostly from formerly unexposed old peat (Moore et al. 2013;Evans et al. 2014;Cook et al. 2018). Water level reestablishment (e.g., ditch blocking) reduces DOC exports in peatlands (Gibson et al. 2009;Armstrong et al. 2010;Wilson et al. 2011;Turner, Worrall, and Burt 2013;Strack et al. 2015), indicating that waterborne carbon loss may be reversed through long-term rewetting and restoration (Evans, Renou-Wilson, and Strack 2016). The understanding of DOC fluxes is complicated during times of elevated peat decomposition and subsidence arising from water level fluctuation, with associated changes in bulk density, vertical hydraulic gradient, porosity, hydraulic conductivity, and pore water residence time (Whittington and Price 2006;Hribljan et al. 2014;Moore, Morris, and Waddington 2015).
The responses of decomposition to water level alteration also vary in fens vs. bogs due to their distinct ecohydrological and biogeochemical conditions (Lafleur et al. 2005;Bridgham et al. 2008;Rydin, Jeglum, and Bennett 2013). A study in Finland showed that the CH 4 fluxes in natural fens were more sensitive to the shift in water levels than those in natural bogs. Specifically, CH 4 emissions reduced by 70% and 45% in natural fens and bogs, respectively, in response to 10 cm water levels decline .
Laboratory microcosm experiments showed that decomposition in peats from fens responded more dynamically to rewetting than those from bogs, and suggested drained fens would experience more DOC leaching after restoration than bogs (Urbanová, Picek, and Bárta 2011). According to modeling studies, decomposition in fens is more sensitive to drawdown than bogs (Gong et al. 2013;Wu and Roulet 2014). The difference in decomposition can be partly ascribed to differences in microbial activity in the two peatland types (Jaatinen et al. 2007).

Indirect effects on decomposition via plant-microbe interactions
Vegetation composition is an essential factor in gas exchange due to the different abilities of functional types to sequester and transport C (Moore et al. 2011;Laine et al. 2012;Korrensalo et al. 2016;Laine et al. 2016;Goud, Watt, and Moore 2018), as well as its interactions with microorganisms (Bardgett and Wardle 2010;Jassey et al. 2014;Robroek et al. 2015). Changes in peatland plant composition influence CO 2 and CH 4 fluxes. For example, as photosynthetic C fixation exceeds C emissions, peatlands dominated by Sphagnum mosses tend to have higher C accumulation rates (Moore et al. 2002;Loisel et al. 2014;Beyer and Höper 2015;Strack et al. 2016;Duval and Radu 2018;Mathijssen et al. 2019). Restored peatlands with more Sphagnum cover often have higher CO 2 uptake and lower variation in CO 2 and CH 4 fluxes Nugent et al. 2018;Swenson et al. 2019). Sphagnum-dominated peatlands also have lower CH 4 fluxes due to the symbiosis with methanotrophic bacteria that metabolize CH 4 to produce biomass and CO 2 (Hanson and Hanson 1996;Frenzel and Rudolph 1998;Blodau, Basiliko, and Moore 2004;Raghoebarsing et al. 2005;Larmola et al. 2010;Kip et al. 2010;Beyer and Höper 2015). Thus, the rehabilitation of Sphagnum after rewetting can lead to a direct GWP benefit and a strong and consistent C sink (Swenson et al. 2019).
In peatlands, the spread of highly productive woody species can benefit C uptake owing to their high rates of photosynthetic C fixation (Ward et al. 2009). Meanwhile, the litter from woody species including fallen branches and dead roots decomposed slowly due to its low surface-area-to-volume ratio and high phenolic content (e.g., lignin) (Blodau and Siems 2012;Gandois et al. 2012Gandois et al. , 2014Dommain et al. 2015;Duval and Radu 2018). Phenolic compounds are toxic to many microbes and restrain microbial extracellular and intracellular metabolism (Fenner and Freeman 2011). They can also bind iron and substrates for microbial growth such as nitrogen and carbon, thus limiting resources of microbial decomposition (Joanisse et al. 2007;Bragazza et al. 2013;Fenner and Freeman 2020). The expansion of shrubs in peatlands eventually increases polyphenol content in plant litter and pore water ) and protects C storage from microbial decomposition during short-term drought (Wang, Richardson, and Ho 2015). Not surprisingly, decomposition rates became faster with the removal of shrubs (Ward et al. 2015). Adding woody litter to peatlands effectively inhibits decomposition during drought, due to leached polyphenolics (Fenner and Freeman 2020). Similarly, Sphagnum produces litter with a high proportion of lignin and water-soluble phenolics, and its exudates are less degradable (Aerts et al. 2001;Bragazza and Freeman 2007;Fenner and Freeman 2011;Leifeld, Steffens, and Galego-Sala 2012;Wang, Richardson, and Ho 2015;Dieleman et al. 2016a;Duval and Radu 2018;Edwards et al. 2018). Under this circumstance, the increased inputs of phenolics and other aromatics derived from expanded woody plants or Sphagnum mosses following water level alteration would help maintain the low oxidation environment, thereby preventing decomposition in peatlands (Couwenberg, Dommain, and Joosten 2010;Dommain, Couwenberg, and Joosten 2011;Hodgkins et al. 2018).
Since plant functional community has strong controls on below-ground communities and processes, indirect effects on C dynamics should be recognized in peatland management of C sinks, rather than simply reflooding to maintain abiotic conditions. For peatland restoration, especially for climate change mitigation, revegetation of key vegetation types for C sequestration (e.g., Sphagnum) is more essential than simply raising water levels, which reinforces the importance of ecohydrologic rehabilitation in future restoration efforts. As Sphagnum vegetation often recovers slowly, introducing fast-growing vascular plant species in the earlier stage of restoration may help rebuild C sinks in peatlands in the short term.
In the meantime, the regulation of plant functional community on microbial activity with altered water levels has been widely discussed, but the study on microbial mediation of plant community dynamics under hydrologic alteration is limited. Because microorganisms play a key role in nutrient mineralization and uptake, their responses to changed abiotic conditions could further impact primary production and overall ecosystem functioning (Geisen 2016;Geisen et al. 2017), which is a situation that might be addressed in future research.

Effects of water level alteration on carbon balance
C balance is changed in many ways by altered water level, with varied results including gains, losses, or no change in C storage. In dried conditions, the labile C pool near the peatland surface is prone to losing C through peat respiration (Chimner and Cooper 2003b). Changes in peat respiration in altered water conditions, especially under drawdown, were the biggest factor in changes in C balance, while plant-driven respiration was less important (Riutta, Laine, and Tuittila 2007;Dorrepaal et al. 2009;Dieleman et al. 2016a;Olefeldt et al. 2017;Jassey et al. 2018).
C balance losses mainly correspond to water level decline, while rising water levels contributes to gains in C balance. In drawdown conditions, reduced C storage occurs as the result of higher ER relative to altered plant production in fens (Schreader et al. 1998;Strack et al. 2006a;Riutta, Laine, and Tuittila 2007;Strack and Waddington 2007;Chivers et al. 2009;Pearson et al. 2015;Olefeldt et al. 2017;Duval and Radu 2018) and bogs (Weltzin et al. 2000(Weltzin et al. , 2001Blodau, Basiliko, and Moore 2004;Munir et al. 2015). In contrast, wet sites accumulate more C than dry or reference sites, mostly because of lower ER (Chimner and Cooper 2003a;Chivers et al. 2009;Helfter et al. 2015;Chimner et al. 2016). A meta-analysis of peatlands in Southeast Asia showed yearly emission rates of 250 g C m −2 for each 10 cm of water level decline in comparison to the normal level (Couwenberg, Dommain, and Joosten 2010). Modeling predicts net losses despite increases in plant production, with an average annual loss of 70 g C m −2 in fens in a scenario of 100-year drying (Chimner, Cooper, and Parton 2002).
Approximate 400 g C m −2 of C was lost in drained peatlands of the Zoige Plateau from 1980 to 2010 (Ma et al. 2016). Wetter conditions occurring as permafrost thaws in peatlands are predicted to create initial C release in the short term (decadal), but yet enhance net C deposition and mitigate climate warming due to relatively higher plant productivity in the long term (centuries) (Wilson et al. 2017).
C balance gains under drawdown are reported more often in boreal peatlands with low decomposition rates, where increased woody production is greater than increased ER (Laiho 2006;Lohila et al. 2007;Strakova et al. 2012;Potvin et al. 2015). As afforested ground vegetation recolonizes, drained peatlands can become a C sink again, absorbing 120-300 g C m −2 year −1 , depending on forest age (Hargreaves, Milne, and Cannell 2003;Meyer et al. 2013;Hommeltenberg et al. 2014). Drainage for forestry in Finland increased the C sequestration rate in peatlands due to the increased growth of shrubs and trees (Minkkinen et al. 1999(Minkkinen et al. , 2002Lohila et al. 2011;Ojanen, Minkkinen, and Penttila 2013), but such drainage may leave peatlands vulnerable to fires, which can rapidly release the C sequestered by aboveground biomass (Flannigan et al. 2001;Turetsky, Donahue, and Benscoter 2011). Turetsky, Donahue, and Benscoter (2011) reported that after long-term experimental drainage, combustion carbon losses during wildfire became ninefold higher in a boreal peatland, even though the carbon accumulation rates have doubled. Low-severity fire modifies the chemistry of soil organic matter and further, stabilizes C storage in peatlands from all regions (Flanagan et al. 2020), but these positive effects may be weakened with increased fire severity after drainage or drought. Indirect effects of water table alteration through wildfire are overlooked and therefore, need to be incorporated in future research.
Similarly, C storage in fens is more sensitive than bogs to environmental change because fens have a narrower tolerance to changes in water level Wu et al. 2012). Modeling studies of C cycling in fens and bogs under most IPCC future scenarios suggest that future bogs may function as C sinks with reduced rates of C absorption, while fens may become C sources (Wu and Roulet 2014).

Concluding remarks
Water level changes induced by climate change and anthropogenic disturbance could substantially influence the status of peatlands as C sinks. Peatland ecosystems including both bogs and fens can maintain their function during water level alteration despite vegetation changes. Water level shifts drive vegetation change particularly because woody plants generally are more productive under drawdown conditions, while Sphagnum mosses are favored in wetter or even flooded conditions. Responses of sedge species to water level alteration vary due to their different responses to hydrological and biogeochemical environments. After peatland drainage or drought, increased peat decomposition is inevitable, but increased plant production may counteract the C loss induced by increased ER. In contrast, rising water levels benefits anaerobic conditions and creates a C sink, unless flooding is associated with a decline in plant production. DOC fluxes in peat are driven by interacting environmental factors, hence making these flux patterns under water level alteration complicated. Bogs and fens differ in their ability to maintain their statuses as C sinks because bogs have a wider tolerance to water level change than fens.
There are many uncertainties regarding vegetation feedbacks on the processes affecting C storage. For example, the feedback cycle of sedge change driving C emission and sequestration is not straightforward. Meanwhile, studies on how vegetation responds to elevated CO 2 (i.e. CO 2 fertilization) related to increased emissions in peatland remains few. The interactions between changed peat properties and water levels are complex, and how these interactions would subsequently influence DOC fluxes are still unclear. Moreover, indirect effects, such as microbial mediation of vegetation dynamics and wildfire under water level alteration are overlooked, and their contributions in ecosystem C cycling should be incorporated in future research. Because most previous studies were conducted in northern peatlands, we encourage comparative studies with other regions of the world. More studies on carbon processes after the restoration of hydrology would be helpful in the discussion of the long-term status of peatlands as C sinks.

Funding
This work was supported by the Regional Innovation and