The role of Phragmites on the CH 4 and CO 2 fluxes in a minerotrophic peatland in Southwest Germany Merit

Peatlands are interesting as carbon storage option, but are also natural emitters of the greenhouse gas methane (CH4). Phragmites peatlands are particularly interesting due to the global abundancy of this wetland plant (Phragmites australis (Cav.) Trin. ex Steud.) and the highly efficient internal gas transport mechanism, which is called Humidity Induced Convection (HIC). The research aim was to (1) clarify how this plant-mediated gas transport influences the CH4 fluxes, (2) which other environmental variables influence the CO2 and CH4 fluxes, and (3) whether Phragmites peatlands are a net 10 source or sink of greenhouse gases. CO2 and CH4 fluxes were measured with the eddy covariance technique within a Phragmites-dominated fen in Southwest Germany. One year of flux data (March 2013 to February 2014) shows very clear diurnal and seasonal patterns for both CO2 and CH4. The diurnal pattern of CH4 fluxes was only visible when living green reed was present. In August the diurnal cycle of CH4 was most distinct, with 11-times higher midday fluxes (15.7 mg CH4 m 2 h -1 ) than night fluxes (1.41 mg CH4 m -2 h -1 ). This diurnal cycle correlates the highest with global radiation, which suggest a 15 high influence of the plants on the CH4 flux. But if the cause would be the HIC, it is expected that relative humidity would correlate stronger with CH4 flux. Therefore, we conclude that in addition to HIC, at least one additional mechanism must be involved in the creation of the convective flow within the Phragmites plants. Overall, the fen was a sink for carbon and greenhouse gases in the measured year, with a total carbon uptake of 221 g C m -2 yr -1 (26% of the total assimilated carbon). The net uptake of greenhouse gases was 52 g CO2-eq m -2 yr -1 , which is summed from an uptake of CO2 of 894 g CO2-eq m -2 20 yr -1 and a release of CH4 of 842 g CO2-eq m -2 yr -1 .


Introduction
Approximately one third of the world's soil carbon is stored in peatlands, although they cover only 3% of earth's total land surface (Lai, 2009). Therefore, peatland conservation or restoration as a climate change mitigation option has recently gained 25 much attention (Bonn et al., 2014). Apart from the positive effect of carbon storage, peatlands are also natural emitters of methane. Methane is a 28-times stronger greenhouse gas than carbon dioxide calculated over a 100-year cycle (IPCC, 2013).
Estimates of methane emissions from peatlands range between 30-50 Tg yr -1 worldwide (Roulet, 2000). There is a high variation in methane emissions. This variability, however, and all underlying processes are not yet well understood (Hendriks et al., 2010;Segers, 1998). It is therefore essential to gain more knowledge about the role of methane in the 30 greenhouse gas budgets of peatlands.
In wetland ecosystems, methane can be transported from the soil to the atmosphere via diffusion, ebullition and via aerenchyma of roots and stems of vascular plants ( Le Mer and Roger, 2001;Hendriks et al., 2010). The largest part of the methane produced in peatlands is directly oxidized in the soil (Le Mer and Roger, 2001;Brix et al., 2001;Lai, 2009). The extent of oxidation depends on the gas transport pathway and is highly dependent on the position of the water table (Le Mer 35 and Roger, 2001;Brix et al., 2001;Lai, 2009) and the presence of vascular wetland plants (Grünfeld and Brix, 1999;Hendriks et al., 2010). Compared to other wetland plants, Phragmites australis (common reed) appears to have a high ability to transport gases between the soil and atmosphere (Salhani and Stengel, 2001).
The gas exchange within Phragmites plants takes place via convective flow through the culm. Currently it is believed that this transport originates from creating a humidity-induced pressure gradient between the internal culm and atmosphere 40 (Armstrong and Armstrong, 1990;Armstrong and Armstrong, 1991;Armstrong et al., 1996b;Afreen et al., 2007). The pores (stomata) in the leaf sheaths of Phragmites are more resistant to pressure flow than against gas diffusion. Due to the higher humidity in the internal culm of the reed, O 2 and N 2 concentrations inside the plant are diluted. Therefore O 2 and N 2 are transported along the concentration gradient from the atmosphere into the sheaths and a higher pressure is created. This causes an airflow from the green living reed stems to the rhizomes and goes back to the atmosphere via dead/broken stems 45 that are still connected to the rhizomes. This mechanism is more than 5 times as efficient as diffusion (Brix et al., 2001) and is also found in other wetland plants (e.g. Nuphar, Eleocharis, Nelumbo and Typha) that have a submerged rhizome system (Dacey and Klug, 1979;Dacey, 1987;Bendix et al., 1994). In a Phragmites dominated wetland, 70% of the produced methane is transported through the plants (Brix, 1989). This means that methane emissions should be highly dependent on this transport mechanism. Apart from this potential influence of HIC on the methane fluxes, Phragmites 50 wetlands can also accrete large amounts of carbon in the soil due to the high annual primary production compared to other wetland plants (Brix et al., 2001;Zhou et al., 2009).
Several studies on methane emissions (Kim et al., 1998a;van der Nat and Middelburg, 2000) and CO 2 emissions (Zhou et al., 2009) from Phragmites-dominated wetlands have been published. Most of them used the closed chamber method.
Despite Phragmites australis being the most abundant wetland species on earth, to date, the eddy covariance (EC) technique has only been used at two study sites: Kim et al. (1998a) performed CH 4 flux measurements in a fen in Nebraska, USA, and Zhou et al. (2009) measured CO 2 fluxes from a Phragmites wetland in Northeast China. To our knowledge, there exist no EC CO 2 and CH 4 flux data from European Phragmites wetlands.
To contribute to a better understanding of the role of Phragmites on CH 4 and CO 2 fluxes, flux measurements were done in the minerotrophic peatland "Federseemoor" located in Southwest Germany. With the eddy covariance method, we were able 60 to measure the net ecosystem exchange of CH 4 and CO 2 in high temporal resolution. This made it possible to detect the influence of the plant-mediated gas transport of Phragmites on the CH 4 fluxes and to evaluate the role Phragmites peatland plays in climate change. We recorded the diurnal and seasonal patterns of these fluxes, evaluated the impact of environmental variables on the fluxes, and determined the carbon and greenhouse gas budgets of this ecosystem. In this paper, we present the results from a measurement period of one year, from March 2013 to February 2014.

Study site
The study was conducted in the Federseemoor (48.092°N, 9.636°E), a peatland with an area of 30 km 2 , that is located in the region Upper Swabia in Southwest Germany. This region is characterized by its moraines and is located on the edge of a high rainfall zone caused by the Alps. Therefore, with a yearly precipitation around 800 mm and an average temperature of 70 7.1 °C, the area is wetter and colder than the average for Germany. The Federseemoor has developed via natural  Grüttner and Warnke-Grüttner (1996)), located in the Federal State of Baden-Württemberg in Germany. The Eddy-Covariance (EC) tower was built northeast of the lake in the center of the largest reed area. Discuss., doi:10.5194/bg-2016Discuss., doi:10.5194/bg- -122, 2016 Manuscript under review for journal Biogeosciences Published: 7 April 2016 c Author(s) 2016. CC-BY 3.0 License.

Biogeosciences
terrestrialisation from a proglacial lake of 30 km 2 that was formed after the last ice age. The lake diminished to a size of 12 km 2 , surrounded by fen and bog. Between the years of 1787 and 1808, the lake size was further reduced by drainage activities to a size of 1.4 km 2 . The resulting 11 km 2 of reclaimed land was meant for agricultural purposes, but appeared to be unprofitable. Natural vegetation started to develop and today it is a nature conservation area, mainly consisting of fen but 75 also containing transitional bog and wooded swamp.
The lake Federsee is completely surrounded by Phragmites vegetation, with a total area of 2.2 km 2 and a density of approximately 70 living shoots per m 2 . To the northeast of the lake, in the middle of the reed, an eddy covariance (EC) tower was constructed (Fig. 1).

80
The location of the EC tower was selected so that only reed vegetation is within the footprint. A LI-7700 open-path CH 4 gas analyser (LI-COR Inc., USA), a LI-7200 enclosed-path CO 2 /H 2 O gas analyser (LI-COR Inc., USA) and a WindMaster Pro sonic anemometer (GILL Instruments Limited Inc., UK) were installed at a height of 6m, twice as high as the reed canopy.
Molar mixing ratio/mass density of the gases and wind speed in three directions were measured at a frequency of 10 Hz.
Air temperature and air relative humidity (HMP155, Vaisala Inc., Finland) and incoming and outgoing shortwave and 85 longwave radiation (CNR4, Kipp & Zonen Inc., The Netherlands) were measured at a height of 6m. Soil temperature was measured in 5, 15 and 30cm depth (LI-COR Inc., USA). Groundwater table was continuously measured with a groundwater datalogger (MiniDiver, Eijkelkamp Agrisearch Equipment Inc., The Netherlands). Rainfall (TR-525USW, Texas Instruments Inc., USA) was measured above the canopy (at a height of 3m). These environmental variables were measured every minute with exception of the water table height, which was measured every 30 minutes. Vegetation height was measured weekly. The declination of the angle-of-attack, caused by the shape of the anemometer, was corrected according to Nakai and Shimoyama (2012). To correct the tilt of the anemometer or angle of the mean horizontal wind, the double rotation method was applied (Wilczak et al., 2001). To convert from CH 4 mass density to molar concentrations, data were compensated for density fluctuations due to changes in water vapour and temperature (Webb et al., 1980). This does not apply to CO 2 /H 2 O gases, since the temperature and pressure are maintained constant in the enclosed path gas analyser. Therefore mixing ratios were used for the flux calculation.

100
The calculated fluxes were checked for quality by means of the 1-9 flagging system of Foken et al. (2004). Only fluxes with quality flags 1-6 were used for further data processing. Outliers were filtered out by removing fluxes that were more than 4 Biogeosciences Discuss., doi:10.5194/bg-2016-122, 2016 Manuscript under review for journal Biogeosciences Published: 7 April 2016 c Author(s) 2016. CC-BY 3.0 License. times the median within a time window of 6 hours and with 6 or more data points within this time window. Because of the often low turbulent conditions and stable stratification during the night, night fluxes with an average friction velocity <0.15 m/s were not considered in the data analysis.

Gap filling
Due to technical failures and discarding data due to flux quality criteria, 46% of the CH 4 data and 35% of the CO 2 data were missing. Gaps were filled with the online tool provided by the Max Planck Institute for Biogeochemistry in Jena (Germany) (http://www.bgc-jena.mpg.de/~MDIwork/eddyproc/). This tool uses the look-up table method described by Falge et al. (2001) and Reichstein et al. (2005). This method was developed to fill CO 2 flux gaps. It uses the correlation of CO 2 fluxes 110 with meteorological variables like global radiation, ambient temperature and vapour pressure deficit.
To date there is no established gap filling method for CH 4 . Nevertheless, we found clear correlations between CH 4 fluxes and global radiation, temperature and relative humidity in our data. Therefore, we used the same gap filling method for CH 4 as for CO 2 .
In the case of power or data logger failure, meteorological data were taken from a meteorological station run by the Federal

115
State of Baden-Württemberg (LUBW) at a 2.2 km distance from the EC station.
Even with these data, the online tool still lacked sufficient data to properly fill a 2 month data gap that was caused by insufficient solar power within the time period 24-11-2013 to 30-01-2014. This was due to the maximum time window (14 days) that the tool uses. Therefore, a look-up table was made manually, to fill this data gap. Global radiation classes with an interval of 100 W m -2 (from 0 to 800 W m -2 ) and ambient temperature classes with an interval of 4 °C (from -10 to 18 °C) 120 were created. Per combined class of temperature and global radiation the average flux was used from the available data from November and February with the same class. Gaps in the look-up table were filled by linear interpolation. Data from the filled gap of December to January are only used for the annual carbon budget estimation, but not for statistical analyses.

Separating NEE
The CO 2 fluxes measured with eddy covariance are the net ecosystem exchange (NEE). By definition, this is the gross 125 ecosystem production (GEP) minus the ecosystem respiration (R eco ). Separating R eco and GEP from NEE is done by using a simplified form of the regression model of Lloyd and Taylor (1994)

145
Biserial (Pearson's) and partial correlation coefficients were calculated to explore the relationship between measured gas fluxes and environmental factors. For this we used the data at half-hourly resolution. Because the samples are autocorrelated in time and hence not independent, no confidence intervals were inferred and correlation coefficients will be interpreted solely in a descriptive manner.
The impact of environmental factors on the gas fluxes was analysed by linear regression. Because of the autocorrelative 150 structure in the data series regression was applied in the framework of the ARIMA (auto-regressive integrated moving average) Box-Jenkins modeling approach. To achieve stationarity (constant expectation and variance) all data (gas fluxes and potential regressor variables) were differenced prior to the analysis. For the ARIMA analysis we used the daily averaged data measured in the vegetation period from 14 May to 31 October.

Seasonal pattern in gas fluxes and environmental variables
The daily averages of CO 2 and CH 4 fluxes are presented in Fig. 3, together with the most important environmental variables.
Only data to the 25-11-2013 are shown, because of the high amount of missing data after this date. The northern 160 hemisphere´s seasonal pattern is clearly visible in temperature and global radiation. These variables show the highest values in July, with average air and soil temperatures of 19 °C and 14 °C, respectively, and daily averaged global radiation of 278 W m -2 . During the whole year, the water table never dropped below the soil surface, which means that the soil was watersaturated all the time. April. From that moment, the reed plants assimilated CO 2 and daily CO 2 fluxes became negative. At the same time CH 4 fluxes rapidly increased. With green reed present, both CO 2 and CH 4 daily fluxes mainly follow global radiation (see Sect. 3.3), but in an inverse manner. This suggests a high influence of the vegetation on both fluxes. The highest CO 2 fluxes were measured in July, the month with the highest temperatures and maximum reed height (260 cm). In July, the average flux was -17.5 g CO 2 m -2 d -1 . The highest fluxes of CH 4 were measured in August, with an average of 0.151 g CH 4 m -2 d -1 . From early

170
October, when the reed entered the senescence stage, fluxes became smaller (CO 2 positive) and on average there was no longer uptake of CO 2 . The lowest fluxes were measured in winter (November to February, data not shown), with an average release of 2.72 g CO 2 m -2 d -1 and 0.044 g CH 4 m -2 d -1 .

Diurnal pattern
To see how diurnal cycles of both CO 2 and CH 4 fluxes change over the season, the monthly averaged diurnal fluxes of both gases are presented in hourly resolution (Fig. 4). CO 2 shows a weak diurnal pattern in the months of March and April, while there is no pattern visible for CH 4 . From May on, when new reed was present, a distinct diurnal pattern was established for both gases, with the highest negative fluxes for CO 2 and highest positive fluxes for CH 4 around noon. The highest midday to night difference for CH 4 was observed in August with, on average, a midday flux of 15.7 mg CH 4 m -2 h -1 and a night flux of  CO 2 m -2 h -1 ). Also in this month the highest night flux was observed, on average a release of 0.629 g CO 2 m -2 h -1 .
The diurnal pattern of CO 2 disappeared in October. From November on, only positive fluxes were measured. The diurnal pattern of CH 4 continued one month longer and almost vanished in November.

Factors affecting the fluxes during growing season
becomes negative when air temperature is partialled out. During the winter period, results differ: CH 4 flux correlates most with soil temperature (r=0.371), followed by water table height (r=0.222) (data not shown).
The correlation table for CO 2 fluxes shows the same pattern, but inverse of CH 4 , except that correlations with air and soil temperature are higher than those of CH 4 .

195
The impact of environmental factors on the daily fluxes of CH 4 and CO 2 fluxes was evaluated by regression analysis in the framework of the ARIMA approach. An ARIMA(0,1,1) model was found to be suited to model the flux time series of both CH 4 and CO 2 (Table 1). Global radiation turned out to be the only regressor with a statistically significant impact (P < 0.05) on the CH 4 fluxes, and global radiation and soil temperature on the CO 2 fluxes. Other environmental factors, such as relative humidity and air temperature, also covary with the fluxes, but their possible impact on the fluxes cannot be determined,

200
because it is screened by the above mentioned variables. After differencing the data, the resulting models for CH 4 and CO 2 , are given by Equations (2) and (3) where  is the differencing operator ((e.g, CH 4,t = CH 4,t -CH 4,t-1 ),  a regression coefficient,  the weight of the moving 205 average (MA) term and e t the residual error term that is assumed to be independently normally distributed (white noise). The : Observed and modeled daily fluxes of differenced data for CH 4 (a), original data for CH 4 (b), differenced data for CO 2 (c) and original data for CO 2 (d). The modeled data are created with an ARIMA(0,1,1) model, with Rg as explaining variable for CH 4 and Rg and T soil for CO2. The error terms with the autoregressive part for the modeled data are not included in these graphs.    Table 1: Model parameters and statistics of the CO 2 and CH 4 ARIMA (0,1,1) models.   September, the contribution of GEP was higher than that of R eco plus CH 4 , resulting in a net carbon uptake during these months.

Carbon and greenhouse gas balance
This uptake more than compensates the net release of carbon in the other months. The yearly CO 2 uptake was 894 g m -2 a -1 and the CH 4 emission was 30 g m -2 a -1 (see Table 2). This leads to a 250 net annual uptake of carbon of 221 g C m -2 by the reed ecosystem, corresponding to 26% of the GEP.

255
CH 4 plays a minor role in the carbon balance, but having a global warming potential of 28 (GWP 100 , IPCC 2013), it heavily affects the greenhouse gas balance (see Fig. 8B). With an uptake of 52 g CO 2 -eq m -2 yr -1 , the ecosystem is a minor greenhouse gas sink (see Table 2).

CH 4 fluxes and plant mediated gas transport
In the period that the above-ground plant parts were alive and green, we observed a distinct diurnal pattern in the CH 4 fluxes.
The highest emission was observed around noon and the lowest during the night. Similar diurnal CH 4 flux patterns from Phragmites-dominated wetlands were reported by Kim et al. (1998b) who used eddy covariance (EC), and by van der Nat et 265 al. (1998) and Grünfeld and Brix (1999) who performed studies with closed chambers. The observed pattern can be explained by the gas transport mechanism within the culm of the Phragmites plants. This mechanism is described by Armstrong (1990, 1991) and Armstrong et al. (1992Armstrong et al. ( , 1996aArmstrong et al. ( , 1996b as humidity-induced convection (HIC).
According to these publications, a convective flow is generated due to an elevated air pressure in the plant stem caused by a humidity gradient (regulated by the stomata) between the inner part of the leave sheaths and the atmosphere. The higher pressure creates an air flow through the entire stem and rhizomes which is vented via old (broken) stems. This process starts after sunrise, is at its optimum in the early afternoon, and decreases until sunset (Brix et al., 2001). During the night, when stomata are closed, gas transport in the stems solely takes place via diffusion. Arkebauer et al. (2001) measured air pressure in stems of Phragmites in the field, and observed the same diurnal pattern as we found in the CH 4 flux data.  found the same pattern with four different wetland plants (incl. Phragmites). They both showed that stem pressure

275
(and convective flow in Brix et al.) correlates with radiation, air temperature and relative humidity. These correlations with HIC were also found in lab experiments (Armstrong and Armstrong, 1991). We found only a strong correlation of CH 4 fluxes with global radiation during the growing season. The correlations we found with air temperature and relative humidity can also be explained by the dependency of these variables on global radiation. It is unexpected that the correlation with relative humidity is not prominent, since this is the driving factor behind HIC. Armstrong and Armstrong (1991) found that 280 convective flow and relative humidity were correlated in a lab experiment with a very low, constant light intensity (4.4 W m -2 ). Sunlight intensity can be more than 200 times higher. When we selected our measured data within the same light intensity range (Rg 3-10 W m -2 ), we found exactly the same negative correlation between CH 4 fluxes and relative humidity as Armstrong and Armstrong (1991). With higher light intensities, however, the correlation vanished. In that same study and in another study (Armstrong and Armstrong, 1990) a correlation was found between photosynthetic active radiation (PAR) and 285 air flow within the plant stem. Radiation can create a temperature difference between the stem and air, this increases the pressure inside the stem compared to the air pressure, which can create a convective flow as well. This phenomenon is called thermal transpiration, but in Phragmites the contribution is believed to be small (Armstrong and Armstrong, 1991;Armstrong et al., 1996a). It also appears that convective flow increases much more with PAR than with infrared radiation (Armstrong and Armstrong, 1990), which speaks against the thermal transpiration hypothesis. The strong correlation 290 between global radiation and CH 4 flux that we observed and the fact that the dominant role of radiation was confirmed in the ARIMA analysis suggests that a mechanism related to stomata control or photosynthesis might play a role in the creation of a convective flow. But the question is still how. Based on our data we cannot give an answer to this question.
We found the highest midday-night difference in the month of August when the reed was fully grown. On average, midday emissions during this month were 11 times higher than at night. This is more than 2 times higher than the highest difference 295 Kim et al. (1998b) found in a Phragmites-dominated marsh in Nebraska. In a lab experiment by Grünfeld and Brix (1999), midday and night fluxes differed by a factor 2.5, which is also much lower than in our study. The reason for this deviation might be the density of the Phragmites plants for which convective flow is expected to be directly proportional. At our site, the density of living green Phragmites plants is almost twice as high (68 m -2 ) as in the prairie marsh in Nebraska (Kim et al., 1998b).
The question remains whether the overall CH 4 flux increases or decreases due to the presence of living green reed. In our data, we found a very clear increase in the daily CH 4 flux after the beginning of reed growth. Soil temperature also increased in the month May, but not in proportion to the CH 4 flux. An increase of the CH 4 flux due to the presence of living reed would be in contradiction to an experiment performed by Grünfeld and Brix (1999). They found a decrease in the CH 4 emissions of 34% with the addition of Phragmites to a submerged organic soil. Their explanation is that methanogenesis is 305 reduced and CH 4 oxidation increased due to the transport of oxygen by Phragmites into the rhizosphere. In a soil without reed, the gas transport would be dominated by ebullition. Transport by ebullition is faster than internal plant transport, so that less of the produced methane is oxidized. Hendriks et al. (2010) found the opposite in a field study with water table differences and vascular wetland plants. A high water table and vascular plants showed higher methane emission than the same soil and water table without vascular plants. Kankaala et al. (2004) found a higher contribution of ebullition to the CH 4 310 flux in a less dense Phragmites shore zone (24 shoots m -2 ) than in a dense area (78 shoots m -2 ). The less dense Phragmiteszone showed threefold higher CH 4 emissions than the denser area. Koch et al. (2014) also found a negative correlation of methane fluxes with Phragmites abundance. Given this negative correlation, the high density of 68 shoots m -2 at our site would suggest that total CH 4 flux would be lower compared to wetlands with lower densities. Our observed yearly CH 4 flux of 30 g m -2 a -1 is in the same range as Kankaala et al. (2004) found (20 to 50 g m -2 a -1 ) with similar dense reed vegetation, 315 and indeed almost 3 times lower than the flux measured by Kim et al. (1998b) in a reed density of only 35 shoots m -2 . So even though our site has a relative low net CH 4 flux, it is likely that plant mediated gas transport during the growing season could lead to higher CH 4 emissions compared to the winter season if in both seasons the ebullition is reduced due to the plant density.

Effect of other environmental factors on CH 4 320
During winter, the daily pattern in the CH 4 fluxes was no longer visible. Dead culms of reed are able to transport O 2 into the soil and CH 4 and CO 2 from the soil to the atmosphere, but only by diffusion (Brix, 1989). During the winter months, correlations of gas fluxes with environmental variables were low. Nevertheless, the highest correlation was with soil temperature. This suggests that soil temperature played the dominant role during this period.
Soil temperature influences microbial activity (Le Mer and Roger, 2001). It also influences respiration, which influences the 325 availability of substrate needed for methanogenesis (CO 2 , acetate) (Christensen et al., 2003). Therefore, an increase in temperature leads to higher emissions.
Water table height is known to have a large impact on CH 4 fluxes (Moore and Knowles, 1989;Aerts and Ludwig, 1997;Grünfeld and Brix, 1999;Updegraff et al., 2001), but only for non-flooded peatlands. In our case, the impact was small because the water table was always above surface level (5-40 cm) so that the soil remained anoxic. Also CO 2 fluxes exhibited clear diurnal and seasonal patterns. The fluxes were mainly influenced by the presence of green plants (high negative correlation with global radiation) and temperature changes. A similar diurnal and seasonal variation was observed in a Phragmites-wetland in North-East China based on eddy covariance measurements (Zhou et al., 2009).

CO 2 flux patterns
They also observed the highest CO 2 uptake in July with -13.6 g CO 2 m -2 day -1 , which is lower than our measured uptake of -335 17.5 g CO 2 m -2 day -1 , and a small release of CO 2 in winter, which is in the same range (2.6 g CO 2 m -2 day -1 vs. 2.7 kg CO 2 m -2 day -1 ) as our observations. The difference in July can be explained by the much higher soil and ambient temperature in the study of Zhou et al. (2009), which resulted in higher R eco relative to the increase in assimilation.
The average Q 10 in the study of Zhou et al. (2009) (Xu and Qi, 2001).

Ecosystem as carbon and GHG sink
The yearly CO 2 uptake was 894 g m -2 a -1 and the CH 4 emission 30 g m -2 a -1 . The CO 2 uptake is almost 4 times higher than in a Phragmites wetland in China (Zhou et al., 2009). The difference could be explained by the lower temperature at our site, so that the respiration rate is lower. Our CH 4 flux is in the same range as at sites with similar Phragmites densities (see 345 above). More in general, northern fens show a wide variation in CH 4 fluxes, from close to zero to 300 g CH 4 m -2 a -1 , depending on temperature, water table and vegetation cover, among others (Lai, 2009;Kayranli et al., 2010). Our site is at the lower range of that spectrum.
Summing up CO 2 and CH 4 fluxes of our ecosystem leads to the net annual carbon uptake of 220 g C m -2 , which is 26% of the gross ecosystem production (GEP). It should be noted that the calculated respiration rate during daytime might be 350 underestimated due to the plant-mediated gas transport. Brix et al. (1996) measured that around noon, 5 times more CO 2 was transported from the soil to the atmosphere by Phragmites plants compared to the early evening. Because daytime respiration is only estimated from its nighttime relationship with soil temperature, respiration could be underestimated and therefore the GEP as well. This would mean that the percentage of the GEP stored in the system would be lower than is given above. It is however hard to say how much lower, since we cannot independently assess the respiration rate during daytime.

355
The carbon uptake at our site is much higher than the 65 g C m -2 (5% of the GEP) measured in the Phragmites-dominated wetland in Northeast China (Zhou et al., 2009) (CH 4 is not considered). But it is only half as high as the uptake estimated for a Phragmites-dominated wetland in Denmark (550 g C m -2 a -1 , 47% of GEP; Brix et al., 2001). The temperature during the growing season was lower than in the Chinese wetland, on average even 10 °C lower in July. Zhou et al. calculated a much higher R eco , which may have caused the difference. The wetland described by Brix et al. (2001) has a similar R eco as ours, but 360 a 30% higher GEP, which explains the diverging findings. Our measured uptake of 65 g C m -2 a -1 fits in the wide range of measured carbon exchange in northern peatlands: from an uptake of 220 g C m -2 a -1 to a release of 310 g C m -2 a -1 (Strack, Biogeosciences Discuss., doi:10.5194/bg-2016-122, 2016 Manuscript under review for journal Biogeosciences Published: 7 April 2016 c Author(s) 2016. CC-BY 3.0 License.
Our CH 4 fluxes show distinct diurnal cycles, but only in the period when living green plants were present. This strongly suggests that plant-mediated gas transport of (a convective transport in the stem of Phragmites) plays an important role regarding the emission of CH 4 from a natural fen site in the Federseeried, Southern Germany. The convective flow within the plant is probably not solely driven by the humidity gradient between the interior of the plant and ambient air (HIC theory).
From our data it is more likely that global radiation plays a more significant role in creating a higher pressure inside the 370 plant.
Our research site is in the measured year a sink for both carbon (-221 g C m -2 a -1 ) and greenhouse gases (-52 g CO 2 -eq m -2 a -1 ). This is probably due to the high productivity of Phragmites plants, high water table and the relatively cold climate so that respiration rates are relatively low. Thereby, the low CH 4 emission compared to other Phragmites wetlands can be explained by the high plant density in our system, which could reduce ebullition.

375
In general, the role of wetland plants that can enhance gas transport, such as Phragmites, is important to consider for the determination of the impact of these wetlands on climate change. The role of environmental factors such as global radiation and relative humidity on the convective flow within Phragmites should be further investigated. This would be helpful to gain more knowledge about the contribution of the plant-mediated-transport to the net fluxes of CH 4 and CO 2 .