The unexpected long period of elevated CH4 emissions from an inundated fen meadow ended only with the occurrence of cattail (Typha latifolia)

Drainage and agricultural use transform natural peatlands from a net carbon (C) sink to a net C source. Rewetting of peatlands, despite of high methane (CH4) emissions, holds the potential to mitigate climate change by greatly reducing CO2 emissions. However, the time span for this transition is unknown because most studies are limited to a few years. Especially, nonpermanent open water areas often created after rewetting, are highly productive. Here, we present 14 consecutive years of CH4 flux measurements following rewetting of a formerly long‐term drained peatland in the Peene valley. Measurements were made at two rewetted sites (non‐inundated vs. inundated) using manual chambers. During the study period, significant differences in measured CH4 emissions occurred. In general, these differences overlapped with stages of ecosystem transition from a cultivated grassland to a polytrophic lake dominated by emergent helophytes, but could also be additionally explained by other variables. This transition started with a rapid vegetation shift from dying cultivated grasses to open water floating and submerged hydrophytes and significantly increased CH4 emissions. Since 2008, helophytes have gradually spread from the shoreline into the open water area, especially in drier years. This process was periodically delayed by exceptional inundation and eventually resulted in the inundated site being covered by emergent helophytes. While the period between 2009 and 2015 showed exceptionally high CH4 emissions, these decreased significantly after cattail and other emergent helophytes became dominant at the inundated site. Therefore, CH4 emissions declined only after 10 years of transition following rewetting, potentially reaching a new steady state. Overall, this study highlights the importance of an integrative approach to understand the shallow lakes CH4 biogeochemistry, encompassing the entire area with its mosaic of different vegetation forms. This should be ideally done through a study design including proper measurement site allocation as well as long‐term measurements.


| INTRODUC TI ON
Natural or pristine peatlands are among the most important soil carbon (C) stocks, despite their relatively small contribution of about 3% to the land mass (Joosten et al., 2012;Leifeld & Menichetti, 2018;Xu et al., 2018). In a pristine state, they have a slightly positive or neutral impact on the Earth's climate system through their function as a strong carbon dioxide (CO 2 ) sink, a weak methane (CH 4 ) source, and a very weak nitrous oxide (N 2 O) source (Frolking et al., 2006;Whiting & Chanton, 2001). Meanwhile, globally about 11% of peatlands (in Europe about 50% and in Germany about 92%) have been drained mainly for land use (Joosten, 2009;Tanneberger et al., 2021;Tiemeyer et al., 2020). These peatlands are strong sources of CO 2 and N 2 O as a result of aeration-induced mineralization processes within the peat body (Joosten, 2009;Joosten et al., 2016;Page & Baird, 2016). Globally, they constitute about 5%-10% of the anthropogenic greenhouse effect (Frolking et al., 2006;Smith et al., 2014;Strack et al., 2022;Tiemeyer et al., 2020). Not least for this reason, efforts have been stepped up worldwide for several decades to eliminate the negative climate effect of peatlands by restoring them through rewetting (Andersen et al., 2017;Bianchi et al., 2021;Humpenöder et al., 2020;Jurasinski et al., 2020). However, it is becoming increasingly clear that the restoration of peatlands, if it is possible at all, will take decades as drainage has often led to: (1) serious loss of typical peatland flora and fauna; (2) severe deterioration of the hydrological and physical peat properties; (3) extremely uneven and heavily subsided peat surface; and (4) sharp increase in the concentration of nutrients in the uppermost peat layers (Klimkowska et al., 2010(Klimkowska et al., , 2019Kreyling et al., 2021). In particular, the restoration of formerly drained, nutrient-rich fens that are completely flooded in the course of rewetting, thus creating shallow lakes or smaller pools with open water areas (Beadle et al., 2015;Beyer et al., 2021), is considered to be potentially problematic. If the surface of the drained fens has dropped below the water level (WL) of the surrounding receiving waters, hydrological measures such as stopping to maintain or filling the drainage system as well as opening the dikes around those areas will lead to flooding of the adjacent fens. Shallow lakes and pools created by such low-cost, low-input dike opening can now be found in many regions with former valley fens (Beadle et al., 2015;Hemes et al., 2019) including Mecklenburg-Western Pomerania in NE Germany. Common to these sites is a broad and dynamic mix of highly productive emergent helophytes such as Phragmites sp., Typha sp. or Carex spp. and open waters (Steffenhagen et al., 2012;Zerbe et al., 2013). Here, the combination of permanently high inundation and contents of mineral and organic matter in the often brownish colored water or in the upper sediment layer is considered particularly critical in terms of net greenhouse gas (GHG) emissions (Zak et al., 2010(Zak et al., , 2019Zak & Gelbrecht, 2007). For example, a high WL (>0.5 m above soil surface) can substantially delay the establishment of CO 2 -fixing helophytes, such as reeds or sedges (Zerbe et al., 2013) as prerequisites for the formation of a strong CO 2 sink. At the same time, high WL can lead to increased fluxes of CH 4 , whose global warming potential is 28 times higher than that of CO 2 (IPCC, 2014).
Due to the high WL, a stable anaerobic zone can form in the surface water above the sediment as well as within the sediment below. The further combination with the high nutrient concentration, the litter of the dead crops or grasses and the increased water temperature in summer due to the dark coloring leads to ideal conditions for CH 4 formation. This may also apply to the absence of emergent helophytes since their aerenchymatic system can promote CH 4 oxidation through the input of oxygen into the sediment or peat close to the roots (Bridgham et al., 2013;Hemes et al., 2018;Knox et al., 2021;Segers, 1998;Whalen, 2005). On the other hand, there is evidence that it is primarily plant litter from grasses and helophytes rather than old peat that serves as a C source for CH 4 formation (Hahn-Schöfl et al., 2011;McNicol et al., 2020;Tuittila et al., 2000). This ultimately means that the period of increased CH 4 emission may continue as long as fresh plant litter remains present and emergent helophytes remain absent. Irrespective of that, the high overall potential to mitigate GHG emissions through rewetting of peatlands is widely accepted . Indeed, there is evidence that CH 4 emissions from shallow lakes formed over formerly drained fens are initially very high, similar to other types of rewetted peatlands (Franz et al., 2016;Koebsch et al., 2015;Wilson et al., 2009).
However, it remains to be clarified how the dynamics of CH 4 release from rewetted peatlands will develop in the longer term or if and when it will level even off to the stage of undisturbed peatlands. The few existing studies on the longer term effects of rewetting on CH 4 emissions were initiated years or decades after the implementation of the restoration measure, often explicitly or implicitly assuming that no changes in CH 4 emissions have occurred in the meantime (Hemes et al., 2018;Kandel et al., 2019;Nugent et al., 2018;Schaller et al., 2022;Vanselow-Algan et al., 2015;Wilson et al., 2016).
The objective of our study was to address and clarify the following questions: (1) What are the dynamics and intensity of CH 4 fluxes after rewetting? (2) How long does it take for measurement site CH 4 emissions to return to the low levels of undisturbed peatlands? (3) Which factors besides the WL or factor constellations determine the development of CH 4 emissions? To fill this knowledge gap, we present the results of 14 continuous years of CH 4 flux measurements of a shallow lake in north-east (NE) Germany, which was created by the rewetting of a formerly drained fen meadow. CH 4 flux measurements were conducted at an inundated and non-inundated site using manual closed chambers. Thus, it allows to measure distinct vegetation types covering the manual chamber plots during the measurement period. Measurements at the inundated site covered K E Y W O R D S cattail, fen, long-term data, methane emissions, peatland rewetting, wetland 1 year prior to and 14 years after rewetting (2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017) while measurements at the non-inundated site covered 12 years (2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015).

| Study site
The study site area was established in a typical percolation peatland in the southern Baltic region (Succow & Joosten, 2001) near the village of Zarnekow in Mecklenburg-Western Pomerania, within NE of Germany. The study area is a part of the Polder Zarnekow-Upost (N53°52.5′, E12°53.3′) situated in the River Peene Valley southeast of the city of Dargun (see map in Figure 1). The entire polder encompasses an area of 550 ha (Schmidt, 2004). Peat thickness is up to 10.2 m. Drainage started during the 18th century. As a result of decades of drainage for intensive cropland or grassland use from the 1970s to the 1990s of the last century (Gelbrecht, 2008), surface subsidence (1 m), strong decomposition, and a shrinking of the upper peat layer (0.3 m) have occurred (Zak et al., 2008(Zak et al., , 2015. The peat layers below had a medium degree of decomposition (Gelbrecht, 2008).
The climate of the study area is moderately continental temperate

Rewetting of the study area was initiated in autumn 2004
through different hydrological measures. At this time, two measurement sites were established, the first at a slightly higher (~0.35 m) landscape position, with an expected minor WL increase ("noninundated site"), and the second within the polder at a lower landscape position, with an expected permanent inundation ("inundated site"). At both sites, WL, sediment (2, 5, and 10 cm depth), and water (5 cm above the sediment) temperatures were measured half-hourly using pressure probes (PDCR1830, Campbell Scientific) and thermocouples (T107, Campbell Scientific), connected to a Campbell science data logger (CR 1000, Campbell Scientific). Precipitation, wind speed, wind direction, air pressure, and air temperature were recorded by the climate station installed at the study area (WXT52C, Vaisala). In addition and to compensate for possible system failure, the daily mean air temperature (2 m height) and the daily sum of precipitation between July 2004 and December 2017 were obtained from the German Weather Service (DWD) in the nearby town of Teterow (25 km southwest).

| Vegetation dynamics
Habitat type mapping for the measurement site was performed in 2004. A vegetation field survey of the entire study area was conducted in August and September 2020 (Table S7). Vegetation patches were visually discerned on an unmanned aerial vehicle (UAV) image taken shortly before and then visited in the field (Table S7). Annual vegetation maps from 2013 to 2020 were made based on Planet Labs satellite images. In total, 99 images of a high temporal resolution, a long temporal coverage, and a 5 × 5 m spatial resolution covering red, green, blue, and near-infrared bands were available. A retrospective classification was conducted assisted by a cluster analysis based on the temporal course of the vegetation indices: normalized difference vegetation index (NDVI) and near-infrared reflectance of vegetation (NIR V ) of each pixel. The resulting clusters were interpreted with the help of aerial or UAV images of higher spatial resolution. The accuracy of this method was assessed for 2020 and 2015 with an overall accuracy of 71% and 78%, respectively.

| CH 4 flux measurements
To measure CH 4 fluxes, five replicate round polyvinyl chloride collars (basal area: 0.194 m 3 ) were inserted next to each other at approx. 7 cm depth in the surface at both sites .  Livingston & Hutchinson, 1995). Each chamber was equipped with two vents at the top to connect evacuated glass bottles (60 mL) for air sampling. The initial sample was taken directly before chamber placement, followed by two more samples every 15 min over a total measurement time of 30 min. The sampled air was analyzed for CH 4 and CO 2 concentrations with a gas chromatograph (GC-14A and GC-14B, Shimadzu Scientific Instruments, Japan; detectors: flame ionization detector for CH 4 , electron capture detector for CO 2 ). To prevent leakage, the collars had a channel for water sealing between the collar and the chamber. To obtain easy access to the collars and to avoid disturbances like triggered ebullition events during the measurements, wooden boardwalks were installed at both sites. where f is the CH 4 flux (CH 4 -C in g m −2 ), M represents the molar mass of CH 4 -C (g mol −1 ), p represents the ambient air pressure (Pa), V is the chamber volume (m 3 ), R is the gas constant (8.314 m 3 Pa K −1 mol −1 ), T designates temperature (K) inside the chamber, A is inside-collar surface area (m 2 ), and dc/dt denotes the linear CH 4 concentration change during the

| CH 4 flux calculation and annual emission estimates
Map of Polder Zarnekow-Upost and satellite images showing the study area with the inundated (red dot) and non-inundated measurement site (green dot); Satellite images and pictures represent conditions at the study area prior to rewetting (2004) and 2, 11 and 14 years after rewetting, respectively. S1 and S2 show sampling points for sediment thickness and carbon stocks in 2011 (Table S5) and 2020 (Table S6). The squared area on S1 indicates approximate location of sediment sampling done by Hahn-Schöfl et al. (2011). Map lines delineate study areas and do not necessarily depict accepted national boundaries. measurement time (s −1 ). For a more robust calculation of CH 4 fluxes, in parallel concentration measurements from all five plots per site (n = 15) were merged into one data set and used for further flux calculation . Prior to this, all measured concentrations originating from air samples flagged during GC-Analysis were removed. In addition, CH 4 concentration outliers within the merged data sets were excluded using sixfold of the interquartile range (IQR; Huth et al., 2018). To account for ebullition events right at the beginning of the measurement period, which may potentially bias the obtained flux, initial CH 4 concentration >5 ppm was also removed. Finally, CH 4 flux measurements were checked for consistency and corrected using parallel measured CO 2 concentrations.
In case of negative CO 2 concentration changes >25 ppm during chamber closure, the measurements were excluded based on the assumption of a positive development of CO 2 concentration during opaque chamber measurements. In total, data processing resulted in a loss of less than 2.5% of all sampled CH 4 concentrations. Annual CH 4 emissions were finally obtained by simple linear interpolation of weekly to biweekly measured CH 4 fluxes .

| Statistical analysis
Annual emissions were checked for normal distribution using the

| Environmental conditions during the study period
Annual air temperature, precipitation, and WL are given in Table 1.
The dynamics of air temperature, precipitation, and WL during the study period are shown in Figure 2a,b. The mean annual temperature during the study period was 9.4°C, the coldest year being 2010 with a mean annual temperature of 7.6°C. An overall trend of a slight increase in annual air temperature during the study period was observed. The highest annual temperature mainly occurred in the last 4 measurement years, exceeding 10°C of mean annual air temperature in 2014 and 2017 (Table 1). During the 14-year study period, the mean annual precipitation was 591 mm and included several, peri- the inundated site are presented in Appendix S1 (Table S1).

| Vegetation dynamics
Before rewetting, the vegetation of the study area was largely dominated by reed canary grass (Phalaris arundinacea) at both sites. In addition, grasses also typical of meadow use such as tufted grass (Dactylis glomerata), tall fescue (Festuca arundinacea), and couch grass (Elytrigia repens) occurred (Huth et al., 2013;Zerbe et al., 2013). As a result of rewetting in October 2004, WL rose above the soil surface, creating a shallow polytrophic lake (Gelbrecht, 2008;Steffenhagen et al., 2012) or an inundated site (Zak et al., 2015). During the same period, the non-inundated site was transformed into a semi-humid meadow where mostly the same dominant vegetation was observed as prior to rewetting.

| CH 4 flux dynamics and annual emissions
Annual CH 4 emissions at the non-inundated site were significantly

| Uncertainty of CH 4 emission estimates
On a single flux basis, CH 4 measurements might be biased due to changes in microclimate or pressure disturbances caused by chamber deployment affecting the gas exchange from the peat surface TA B L E 1 Environmental variables and annual methane (CH 4 ) emissions during the study period. Mean annual air temperature (°C), cumulative annual precipitation (mm), and mean annual WL (m) are given for both measurement sites. Dominant vegetation coverage and cluster groups are added for the inundated site.
Year Unit  hydrophytes (Steffenhagen et al., 2012;Zerbe et al., 2013). It is also a common characteristic that the extent of these areas shrinks over the course of time, that is, within years to decades, because of the colonization with diverse forms of emergent helophytes.

Non-inundated inundated
There are contradictory estimations about the future development of vegetation on flooded fens like the Zarnekow-Upost polder.
On the one hand, the formation of cattail could mark the final vegetation stage of the shallow lake before silting up (Zak et al., 2015).
On the other hand, there is also evidence that subsequently reed or sedge meadows might establish (Steffenhagen et al., 2012;Zerbe et al., 2013). This is probably determined by nutrient supply in addition to water level development (Schulz et al., 2011). However, the time span of this development is fairly unknown. In addition to dense Typha stands, a relatively large share of open water areas still remained (>30% of the total area) at the study area even 5 years after the end of the CH 4 flux measurements and 18 years after the rewetting of the area.

| Response of actual annual CH 4 emissions to environmental variables
Environmental conditions such as WL and temperature often explain the dynamics of CH 4 emissions in peatlands (Moore & Dalva, 1993;Morin, 2019;Strack et al., 2004;Turetsky et al., 2008). thus CH 4 production by methanogenic archaea, while higher temperatures generally increase microbial activity (Hopple et al., 2020;Turetsky et al., 2008;Zinder, 1993). A significant positive dependence of annual CH 4 emission with WL (p-value <.1) was also ob- with mean annual air temperature could be seen (Figure 4b).
In summary, mean WL is of central importance for the CH 4 emission potential (Evans et al., 2021;Tiemeyer et al., 2020); however, its effect is modified not only by the temperature driving the The variance in-between transitional stages is explained by environmental conditions (PC1; e.g., WL and CH 4 ), while the variance within a transitional stage is largely explained by weather conditions (PC2; e.g., precipitation and air temperature).
interannual variability, but, more importantly, the vegetation cover, which might alter microclimatic conditions, such as the water temperature, through shading and determines the magnitude of CH 4 emissions ( Figure 4). This is substantiated by the performed LMM ANOVA, which showed a significant effect of WL (p-value = .011) on annual CH 4 emissions. Figure 5 shows the 1:1 agreement plot of annual estimated (interpolation of manual chamber measurements) and predicted (using LMM) annual CH 4 emissions. No correlation could be observed for CH 4 emission with dissolved organic carbon (DOC) concentration measured either in pore or in surface water (Table S1).

| Cause of the extremely high CH 4 emissions, their long duration and decline
Annual CH 4 emissions at the inundated measurement site were much Moreover, in a mesocosm study with rewetted peat, the presence of Typha sp. reduced the CH 4 emission rate to less than one tenth of the unvegetated control (Vroom et al., 2018). Decomposition experiments showed that Ceratophyllum litter was converted to CH 4 up to three times more than litter of emergent helophytes such as T. latifolia or P. australis (Zak et al., 2015). Consistent with this, EC measurements of CO 2 exchange over the open water surface performed by Franz et al. (2016) and  from 2013 to 2017 showed net CO 2 losses, that is, no CO 2 -C accumulation in the form of new sediment. However, sediment and peat analyses carried out in 2020 showed that considerable C accumulation rates occurred after flooding in the open water surface, which were much higher than in the areas covered by emergent helophytes (Figure 1 (S2), Table S6). This can only be explained by ongoing lat- However, the question of how much these two C sources contributed to CH 4 formation can only be answered with the help of targeted F I G U R E 5 1:1-agreement plot between interpolated (measured) and predicted (LMM) annual methane (CH 4 ) emissions at the inundated and non-inundated site. The LMM consists of a quadratic fixed term for WL and a linear random (intercept and slope) term for temperature (Figure 4). Symbols are color coded and shaped according to the different transitional stages. The black line shows 1:1 agreement and the dashed gray line the correlation between interpolated and predicted annual CH 4 emissions. Error bars indicate the estimated error (interpolated CH 4 emissions) and the standard error from the fitted LMM (predicted CH 4 emissions).  Zak et al. (2015) revealed that the CH 4 production potential by T. latifolia was three times lower compared to C. demersum despite the substantially higher annual biomass production of T. latifolia. This is worth mentioning as helophytes in general might increase the input of fresh organic material and thus accelerate anoxic decomposition and CH 4 production (Chanton et al., 1993;Tuittila et al., 2000). Second, the aerenchymatic tissue of T. latifolia creates a transport pathway for oxygen into the anoxic sediment layers, promoting the oxidation of CH 4 by methanotrophs and inhibiting methanogenesis (Fritz et al., 2011;Vroom et al., 2018). However, at the same time, the plant aerenchym might also act as a direct diffusion pathway for CH 4 from anoxic sediment layers into the atmosphere (Chanton et al., 1993;Vroom et al., 2018). Finally, regarding the third and fourth points, the dense stands of helophytes likely lower the temperature in the sediment and water as a result of shading (Franz et al., 2016) and might furthermore hamper the transport of allochthonous fresh organic material into the measurement site.

The increased annual CH
The first is in alignment with an observed drop (for 0.3°C) of the average annual soil temperature (10 cm depth; standardized as difference to average annual air temperature) from 2014 and 2015 compared to 2016 and 2017 at the measurement site.
As shown so far, the different transitional stages following rewetting evidence different but high annual CH 4 emissions (see Section 4.2) that only started to decrease during the last two measurement years after establishment of emergent helophytes (e.g., Typha sp., Carex sp.). However, given the uncertainties regarding the further plant succession, the future CH 4 dynamics at the inundated measurement site remain speculative and thus need further investigation.

| CON CLUS IONS
Thanks to the long-term measurements of CH 4 emissions and veg- and Carex sp. could also lead to a reduction of the extremely high CH 4 emissions (e.g., Couwenberg et al., 2011), these should also be included in future studies.
Last, but not least, the study also suggests that it is quite important for the understanding of the CH 4 biogeochemistry of the shallow lakes that an integrative approach is taken, encompassing the entire area with its mosaic of different vegetation forms. An example of this is the apparently great importance of plant litter from former grass vegetation and the surrounding highly productive helophyte belt for the high CH 4 fluxes of the open water areas. Accordingly, harvesting of grass before rewetting could lower initial high CH 4 emissions. If operable drainage systems with weirs and, perhaps, a pumping station still exist, a more controlled and progressive "slow rewetting" strategy accompanied by the rapid coverage of the entire area with helophytes is proposed as an alternative to spontaneous inundation of long-term drained peatlands (Zak & McInnes, 2022).

ACK N OWLED G M ENTS
The authors thank Bertram Gusovius and Matthias Lück for GC analyses, Natalia Pehle, Elisa Albiac Borraz, and Nicole Jurisch for data preparation, and Hans-Joachim Schröder for in-situ measurements.
Open Access funding enabled and organized by Projekt DEAL.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data of the study can be accessed at doi: https://doi.org/10.4228/ zalf-f1sj-0156.