Time Since Rewetting De nes Vegetation Composition and Carbon Dioxide Fluxes on Former Milled Peatlands-Comparison With Undisturbed Bogs


 Rewetting is the most common restoration approach for milled peatlands in Europe, with the aim of creating suitable conditions for the development of peatland specific plant cover and carbon accumulation. Therefore, it is important to determine if time since rewetting is pivotal for milled peatlands to become functionally and structurally similar to their undisturbed counterparts. We investigate the temporal succession in rewetted peatlands in Estonia by a chronosequence of 4, 15, and 35 years before the measurements. Plant functional type (PFT) cover and biomass, bryophyte production and CO2 fluxes were measured on two milled peatlands, as well as undisturbed bogs adjacent to milled peatlands. Differences in vegetation composition and CO2 fluxes between the sites were greater for rewetted than undisturbed sites. The most recently rewetted site was mainly covered in bare peat and Eriophorum vaginatum and was a CO2 source. On the rewetted site of 15 years, Sphagnum was present in addition to ombrotrophic sedges, and in the rewetted site of 35 years, lawn-hollow microtopography is starting to develop with various PFTs. Both of these sites were CO2 sinks. Lawn Sphagnum was abundant on the two older rewetted sites, and was connected with CO2 sink functioning in the rewetted sites. Still, hummock Sphagnum species, which were present in undisturbed bogs, were absent from all of the rewetted sites. With time, CO2 fluxes, microtopography and vegetation develop after rewetting in the direction of undisturbed bogs, while vegetation composition still differs from the reference sites even 35 years after rewetting.


Introduction
Undisturbed peatlands are important carbon sinks in the long term (Yu 2012) as well as a suitable habitat for plant species that have adapted to survive in acidic and waterlogged conditions (Minayeva 2008).
Northern peatlands have been widely affected by peat mining for horticulture or energy production (Leifeld et al. 2019), particularly since the 1950s when peat milling became the main technique for peat extraction. This method involves drainage and the removal of vegetation in large areas, so thin layers of peat can be extracted every summer season.
Excavated peatlands have several negative environmental effects, such as peat loss through mineralisation, high CO 2 emissions, re hazard, no plant diversity and low aesthetic value. The natural revegetation of those site takes a long time, depends on the environmental conditions of the site and usually does not lead to mire-speci c plant communities ; Graf et al. 2008;Orru et al. 2016). Unrestored milled peatlands are important CO 2 sources to the atmosphere due to the low water tables allowing peat mineralisation and sparse or absent vegetation (Strack et al. 2016; Rankin et al. 2018). The main mitigation possibility for those negative impacts is peatland rewetting, which through higher water tables creates suitable conditions for revegetation and thus reduces CO 2 emissions (Wilson et al. 2016) and the ammability of these sites (Granath et al. 2016). Although rewetting increases CH 4 emissions on restored peatlands, but as CH 4 is a short-lived gas in the atmosphere, rewetting of peatlands mitigates the climate change in long-term (Günther et al. 2020). Various criteria have been taken into account throughout the history of restoration ecology to assess restoration success. First, biodiversity measures and hydrology were engaged to indicate the success of restoring the ecosystem, which in recent decades have been integrated with greenhouse gas balances showing recovery of ecosystem functioning (Kløve et al. 2017; Renou- Wilson et al. 2018).
The initial response of the plant community and its diversity to rewetting is complex and depends also on the pre-rewetting state of the peatland (Tuittila et al. 2000). Over the time-scale of several decades after peatland rewetting or self-recovery, plant cover increases with time (Orru et al. 2016; Priede et al. 2016).
Furthermore, CO 2 uxes change in time after rewetting (Kløve et al. 2017). Beyer and Höper (2015) estimate based on their experience of greenhouse gas measurements in temperate peatland that rewetted peat extraction sites may become peat accumulating ecosystems after about 30 years. Even after 30 years post rewetting, milled peatlands can remain CO 2 sources, but those emissions from rewetted sites tend to be smaller than from active peat extraction sites, especially if Sphagnum is dominating (Samaritani et  . Some studies report that vegetation in rewetted sites is more heterogeneous than in pristine peatlands due to the patchiness and incompleteness of the plant cover on restoration sites as vegetation cover starts to develop near the drainage ditches and close to existing vegetation (Soini et al. 2010;Laine et al. 2016 (Kivimäki et al. 2008) and soil organic matter accumulation (Andersen et al. 2013) have been recorded on plots with mixed graminoid and Sphagnum patches compared to pure graminoid patches. Sphagnum species have lower photosynthetic capacities (Korrensalo et al. 2016) and lower respiration, therefore plots with Sphagnum are larger CO 2 sinks than plots with only graminoids (Beyer and Höper 2015). Vascular plants are also important in peatland CO 2 exchange, especially by mitigating the effect of drought on CO 2 sink functioning (Kuiper et al. 2014 In this paper we analyse the success of rewetting by comparing CO 2 uxes and vegetation on relatively similar rewetted milled peatlands with somewhat different site conditions in different successional stages and initially eco-hydrologically similar nearby undisturbed bogs to assess if a longer time since rewetting or ecosystem recovery time leads to ecosystems that functionally converge to the state of reference bogs. For that, we established the following postulates: 1. CO 2 uxes and vegetation structure on rewetted milled peatlands develop in the direction of undisturbed reference bogs in time; 2. Higher amount of plant above-ground, especially Sphagnum, biomass is related to higher CO 2 sink function on rewetted milled peatlands.

Study sites
Two paired study sites (Kõrsa and Hara) were selected ( Table 1), both of which include rewetted abandoned milled peatlands and remnant open raised bog areas (Fig. 1). The Kõrsa site is located in southwestern Estonia next to an active peat extraction site. The Kõrsa rewetted (Kõrsa R ) site has selfrecovered after the end of peat extraction due to the water level being raised up to the peat surface in 1980 following a damming to create a rewater reservoir (Ramst et al. 2007). Revegetation began at The respective paired sites are located in the same mire-basin, and the distances between the rewetted and undisturbed sites range from 140 metres in Hara to 500 m in Kõrsa.
In all of the study sites, four measurement plots per site were established during the previous year (2015).
As two of the plots in Hara RN became ooded during the measurement period, they were omitted from the study, and data from two measurement plots in Hara RN were used. The locations of the permanent measurement plots were chosen based on the dominant vegetation and by taking into account its variability between micro-topographic levels. In undisturbed sites and Kõrsa R , where microtopography has already developed, two measurement plots were situated on the hummocks and two plots on the lawns of each site. In other sites, two replicates for each vegetation type were established. We measured the plant species coverage (%), vascular plant leaf area index (LAI vasc ; m 2 m − 2 ), aboveground biomass of PFTs (AGB; g dm − 2 ) and moss production (AGP; g dm − 2 year − 1 ) as well as the length increment of mosses (LI; mm year − 1 ). The plant cover of measurement plots was determined visually at the peak of the 2016 growing season (end of July) from inside the CO 2 ux measurement collars (four plots per each site/management type combination, but two plots in Hara RN ). LAI vasc was determined according to Wilson  inserted to about a 20 cm depth, with the rim lled with water to ensure an air-tight t during ux measurements. We measured CO 2 concentrations with the infrared gas analyser Li-6400 (Li-Cor (USA)) from transparent Plexiglas chamber (60×60×30 cm) with a cooling system. The measurements period was two minutes, and the CO 2 content in the chamber was recorded with an interval of 15 s. After measuring the CO 2 concentrations in full-light, NEE was measured on two lower irradiation levels by using one or two shades that reduced the photosynthetically active radiation (PAR (µmol m − 2 s − 1 )) reaching the vegetation in the chamber at an average of 65% and 88%, respectively. Lastly, R ECO was measured by covering the chamber with an opaque hood. Between each measurement period, the measurement chamber was ventilated. During the measurement campaigns, plant parameters for determining LAI vasc inside the measurement collars were measured according to Wilson et al. (2007a) in addition to recording PAR, the temperature inside the chamber, peat temperatures at 5 cm and 15 cm depths and the water table (cm).
Input data (PAR, T AIR ) for CO 2 ux reconstruction were measured with hourly intervals in stations belonging to the Estonian Weather Service. For Hara, the temperature data was obtained from the nearest station in Vanaküla (about 10 km from the site) and radiation data from Harku meteorological station (about 70 km from Hara). For Kõrsa, all meteorological data was obtained from Pärnu-Sauga meteorological station located about 15 km from the site. Those stations were the closest to the study sites where PAR and T AIR were continuously measured, and they were located within a 10 km distance from the sea similarly to the study sites.
The ux rates were estimated based on linear change in CO 2 concentrations in time. The linear method was chosen, as this method was considered suitable by Kandel et al. (2016) for CO 2 ux calculations in the case of short (few minutes) chamber closure periods (2 min in current study). The measured NEE and R ECO uxes were considered suitable according to the following quality criteria: variation of PAR during the ux measurement not exceeding ± 15%, variation of inside temperature of the chamber not varying more than ± 5°C and the determination coe cient (R 2 ) of the measured ux of at least 0.9. Very small uxes (± 0.2 ppm s − 1 ) were accepted regardless of their R 2 value. Similar quality criteria in respect of R 2 values were used by Järveoja et al. (2016). A total of 215 CO 2 ux measurements ful lled the set criteria and were used for CO 2 ux reconstructions. Photosynthesis (P g ) was calculated by adding R ECO to NEE. CO 2 uxes were reconstructed for the period from the beginning of May until the end of September 2016 at each site. With these reconstructions, based on measured and calculated CO 2 uxes and other parameters (PAR, LAI vasc and air temperature (T AIR )), models were created for relating differences in measured CO 2 uxes with differences in input parameters for reconstructing the whole growing season The gross photosynthesis (P g (mg CO 2 m − 2 h − 1 )) model uses the saturating response to PAR (Eq. 1) and records the change in LAI vasc during the vegetation season: where P max is the maximum photosynthesis at light saturation, k and s are respectively the PAR and LAI vasc values when P g reaches half of its maximum level.
The respiration model (Eq. 2) expresses an exponential response of ecosystem respiration (R ECO (mg CO 2 m − 2 h − 1 )) to the temperature inside the chamber (T AIR ).
Where parameters r0 and b are respectively the respiration at the 0°C temperature and the sensitivity of respiration to air temperature, and T AIR is the air temperature (°C). CO 2  The multivariate analysis methods Redundancy Analysis (RDA) and Detrended Correspondence Analysis (DCA) were applied in PC-ORD ver. 7 to relate the abundances of PFTs and CO 2 uxes on rewetted and undisturbed sites, and to analyse the changes in those variables with time since rewetting, respectively. In RDA, the response variables were standardised and a randomisation test was applied to test for any signi cant relationship between the PFT and CO 2 ux matrices. DCA was used to nd the main gradients in PFT and CO 2 ux data using time since rewetting and the site as supplementary variables.

Vegetation
Vegetation varied signi cantly between rewetted and undisturbed sites and between all rewetted sites, while small differences also occurred between both undisturbed sites (Fig. 2). More PFTs were present on undisturbed and older rewetted sites, while many PFTs such as Sphagnum and evergreen shrubs were absent from the recently rewetted Hara RS . Evergreen shrubs such as C. vulgaris and A. polifolia had higher cover in undisturbed sites, while V. oxycoccus was present with low cover only in Kõrsa R . Evergreen shrub biomass was absent or signi cantly lower in rewetted sites compared to undisturbed sites (Appendix S1). Ombrotrophic forbs R. chamaemorus and D. rotundifolia were only present in undisturbed plots, but with relatively low cover (0.5-3%). Only in Kõrsa R minerotrophic forbs like Melampyrum species and T. palustris were present. Tree seedling of Salix spp., Betula spp. and P. sylvestris had about 1% cover on all sites, or were absent.

Carbon dioxide uxes
Measured NEE and R ECO varied spatially to a larger extent in rewetted rather than in undisturbed sites (Appendix S2). A higher CO 2 net uptake with higher PAR was measured on both rewetted and undisturbed sites.
Reconstructed P g and NEE did not differ statistically signi cantly (p > 0.05) between the rewetted and the undisturbed sites, whereas R ECO was signi cantly higher in the rewetted than in the undisturbed sites (p < 0.05; Fig. 3). The respiration model's parameter r0 was signi cantly (p < 0.05) higher in the rewetted (41.8 ± 11.4 mg CO 2 m − 2 h − 1 ) than in the undisturbed sites (9.4 ± 3.1 mg CO 2 m − 2 h − 1 ; Appendix S3). In Kõrsa R the reconstructed P g was signi cantly higher than in the undisturbed sites and at Hara RS (p < 0.05). Also, Kõrsa R had signi cantly higher R ECO than Kõrsa P , whereas all of the other sites had a similar R ECO .
Although there were no differences in the model parameters between Hara RN, Hara RS and Hara P (p > 0.05), P max and r0 were higher in Kõrsa R than in Kõrsa P (p < 0.05) indicating a higher maximum CO 2 uptake in case of light saturation and also a higher minimum respiration rate in rewetted sites. Undisturbed sites did not differ signi cantly according to their CO 2 uxes (p > 0.05). There were no signi cant differences in CO 2 uxes between the hummocks and the lawns in the undisturbed sites and Kõrsa R (p < 0.05).
GLMMs were used specify the effect of site, microtopography and time since rewetting on growing season CO 2 ux components (R ECO , P g , NEE) on rewetted peatlands. Although none of the xed effects and GLMMs were statistically signi cant, time since rewetting had strongest effect on all of the CO 2 ux components ( Table 2). In addition to time since rewetting, microtopography and combination of microtopography and site had also relatively strong, but still statistically insigni cant effect on R ECO .

CO 2 uxes and vegetation
CO 2 uxes correlate with every PFT differently between the undisturbed and rewetted plots (Fig. 4, Appendix. S4). In the undisturbed sites, P g was higher in measurement plots with higher ombrotrophic sedge (E. vaginatum) cover and biomass but lower with higher tree cover, which was related with the higher values of the parameter k indicating the PAR value when P g reaches half of its maximum value. In the rewetted sites, NEE was higher in the case of higher Sphagnum abundance, and higher photosynthesis rates were connected with the cover of minerotrophic forbs. P g increases with higher bryophyte and vascular biomass in rewetted sites, whereas this correlation was insigni cant in the undisturbed sites (Fig. 5). In the undisturbed sites, higher R ECO was measured on plots with higher vascular plant biomass, whereas this correlation was insigni cant in the rewetted sites. There were no other statistically signi cant correlations between vascular plant, bryophyte and plant biomass, and P g , R ECO and NEE in the rewetted nor in the undisturbed plots.
With time since rewetting, communities evolve in the direction of undisturbed mires, where several PFTs are present, including Sphagnum and evergreen trees (Fig. 6) years after restoration. Throughout the study period in Kõrsa R and Hara RS , the water level stayed inside the moss layer, mainly near the moss surface, therefore not decreasing the moss growth during the summer period. In the rewetted sites with thick moss layer in the current study, the moss layer was looser than in the undisturbed reference sites. This was probably due to the higher water table along with the high abundance of hollow Sphagna in the rewetted sites. Hollow Sphagnum could be affected from extreme droughts to a larger degree due to their larger pore size and less connectivity with the residual peat layer (McCarter and Price, 2015) than the denser Sphagnum cover of undisturbed bogs, therefore making CO 2 exchange on rewetted sites more susceptible to drought impacts. Some PFTs were lacking or had very low abundances in the rewetted sites but were present in the reference sites. We found signi cantly lower biomass and cover of evergreen shrubs on the rewetted than in the undisturbed sites, similar to results by Soini et al. (2010) and González et al. (2013), and they were absent from the most recently rewetted sites. Hummock Sphagna, which was present in both undisturbed bog sites was completely absent from the rewetted sites. The low occurrence and dying-off of hummock Sphagnum due to high water tables has been reported previously by Soini et al. (2010) and González et al. (2013). In reverse, Karofeld et al. (2015) recorded relatively high cover of hummock Sphagna and the presence of shrubs on restored milled peatland site where those species were dispersed using the mosslayer-transfer technique . Therefore, the application of this technique could lead to a more diverse vegetation composition of restoration sites.
While vegetation differs signi cantly between the rewetted sites, being more diverse in the older sites, the vegetation in both undisturbed sites with a similar hummock and hollow vegetation pattern did not differ from each other. Hummocks on the two undisturbed sites are typical Calluna-vulgaris-Sphagnum fuscum communities, the most common plant associations in Estonian bogs (Masing 1982), and are comparable to the high hummock communities described by Korrensalo et al. (2018). Lawns in the undisturbed sites belong to the tussocky Eriophorum community or the Sphagnum balticum-Sphagnum rubellum community (Masing 1982 Sphagnum has been considered a keystone genus of peatland restoration (Rochefort 2000). In the newly rewetted Hara site, Sphagnum was not yet present in the measurement plots, although some patches of lawn Sphagnum (mainly Sphagnum cuspidatum) were present in depressions with high water level. After rewetting, the height of the water table should remain a few centimetres below the peat surface, which leads to optimal conditions for Sphagnum growth and peat accumulation (Beyer and Höper 2015). Sphagnum has relatively high immigration potential ) and is abundant on the undisturbed plots bordering the rewetted ones, so further colonisation of Sphagna in recently rewetted sites is expected. In both older rewetting sites, Sphagnum had almost total cover. In addition, in the oldest Kõrsa R site, lawn Sphagnum species have created some relatively high hummocks and overgrow E.
vaginatum tussocks. The AGP and IL of Sphagnum in the rewetted sites was similar to those reported by Ilomets (1982) in Estonian undisturbed peatlands, while we measured about double the production and somewhat higher IL of Sphagna on the undisturbed sites. This probably results from different methods

Carbon dioxide uxes
Both the undisturbed sites and the older rewetted sites were CO 2 net sinks during the growing season, while the more recently rewetted site was still a CO 2 source. Variations in CO 2 uxes between the rewetted sites are large due to differences in vegetation, weather and water levels -while some sites are important . Therefore, it could be expected that the CO 2 sink function will increase and be more stable with secondary succession after rewetting, especially as the actual acrotelm is formed with time.
We detected some effect of site status on plant above-ground biomass, which on rewetted sites had a (2019b) also reported interaction between peatland management (undisturbed, rewetted), PFTs and carbon sequestration. According to Järveoja et al. (2016), those correlations depend on water level depth -if the water level is high in restored milled peatlands, bryophyte cover correlates with NEE, P g and autotrophic respiration, whereas with deeper water table CO 2 uxes correlated with vascular plant cover.
Therefore, the different correlations on rewetted and undisturbed sites are consistent with previous studies (Strack et al. 2016) and could be related to differences in water table height and uctuations on sites with different management.
There are large differences in photosynthetic capacities between PFTs. In the undisturbed sites, we measured higher photosynthesis and maximum photosynthesis rates (P max ) in the case of higher E.  ).
In the rewetted sites, higher photosynthesis and P max were measured with higher evergreen shrub cover.
Evergreen shrubs stand out from other vascular plants with low photosynthesis and respiration rates (Laine et al. 2016), while in reverse Korrensalo et al. (2016) reported high maximum photosynthesis rates on evergreen shrubs like A. polifolia, C. vulgaris and V. oxycoccus, which are also present in the undisturbed sites and Kõrsa rewetted site in our study. According to Korrensalo et al. (2016), the P max of evergreen shrubs varies between species belonging to the same PFT. Still, the cause of controversies between different studies remains unclear and can be result of a rather low number of measurements that do not cover the whole ecosystem variation.
High photosynthesis in the case of higher evergreen shrub cover in this study could also be connected with higher plant cover and the number of PFTs on the measurement plots in Kõrsa R where evergreen shrubs were present. According to Kivimäki et al. (2008), the presence of different PFTs lowers the R ECO /P g ratio, so creating conditions for higher CO 2 net uptake as in Kõrsa, while in monostands of E.
vaginatum this ratio is higher, which also explains a lower CO 2 net uptake, as well as CO 2  exchange and biomass related with scarce tree cover were not accounted for in any of the studied sites.
In addition, uncertainties related to ux measurements and reconstructions could affect the source or sink function of the sites during the growing seasons, especially if uxes are very low and uncertainties higher (Bubier et al. 1999).
Third, the methane emissions, along with dissolved organic carbon and dissolved inorganic carbon, were not measured from the study sites in this paper, as the general aim of the study was to analyse the differences in plant production parameters and PFT composition closely related with the CO 2 uxes. Therefore, the results presented here do not provide information about the full carbon balance of the sites, as methane emissions for such sites have been reported to be high (Strack et al. 2014(Strack et al. , 2016 Vanselow-Algan et al. 2015; Beyer and Höper 2015; Günther et al. 2017). Within these limitations, we still hope the paper will be of interest for a wide audience of peatland ecologists.

Conclusion
Although vegetation structure on rewetted milled peatlands approaches this on reference sites with time, some plant functional types present in the undisturbed reference sites, e.g., shrubs, colonise these sites in the later development stages and hummock Sphagnum could be absent even decades after rewetting. Vegetation composition developing with time affects the carbon accumulation of rewetted sites. During the studied growing season, over a decade ago rewetted milled peatlands were carbon sinks similarly to the reference sites, whereas the most recently rewetted site was still a carbon source to the atmosphere.
Although graminoids play an important role in the photosynthesis of rewetted sites, as they do in undisturbed reference bogs, the carbon accumulation of rewetted peatlands is related with development of the Sphagnum mat, which is present in the reference sites. A well-developed Sphagnum mat also re ects the development of other environmental variables, of a functioning acrotelm and the development of a C sink function. Thus, a well-developed Sphagnum lawn could be used as an indicator of successful restoration. However, general plant functional type composition can still differ from reference sites in some accounts even several decades after rewetting.

Conclusion
Although vegetation structure on rewetted milled peatlands approaches this on reference sites with time, some plant functional types present in the undisturbed reference sites, e.g., shrubs, colonise these sites in the later development stages and hummock Sphagnum could be absent even decades after rewetting. Vegetation composition developing with time affects the carbon accumulation of rewetted sites. During the studied growing season, over a decade ago rewetted milled peatlands were carbon sinks similarly to the reference sites, whereas the most recently rewetted site was still a carbon source to the atmosphere.
Although graminoids play an important role in the photosynthesis of rewetted sites, as they do in undisturbed reference bogs, the carbon accumulation of rewetted peatlands is related with development of the Sphagnum mat, which is present in the reference sites. A well-developed Sphagnum mat also re ects the development of other environmental variables, of a functioning acrotelm and the development of a C sink function. Thus, a well-developed Sphagnum lawn could be used as an indicator of successful restoration. However, general plant functional type composition can still differ from reference sites in some accounts even several decades after rewetting.

Declarations
Funding: No funding was received for conducting this study. Average cover (in %) of vascular plant (a) and bryophyte (b) plant functional types of the study sites (±SE). Different small case letters indicate statistically signi cant differences between the sites (p < 0.05). Statistical signi cance was tested using the Kruskal-Wallis test, and pairwise comparison was concluded using the Mann-Whitney test with Bonferroni correction.  Supplements.docx