Flooding-related increases in CO 2 and N 2 O emissions from a temperate coastal grassland ecosystem

Given their increasing trend in Europe, an understanding of the role that flooding events play in carbon (C) and nitrogen (N) cycling and greenhouse gas (GHG) emissions will be important for improved assessments of local and regional GHG budgets. This study presents the results of an analysis of the CO2 and N2O fluxes from a coastal grassland ecosystem affected by episodic flooding that was of either a relatively short (SFS) or long (LFS) duration. Compared to the SFS, the annual CO2 and N2O emissions were 1.4 and 1.3 times higher at the LFS, respectively. Mean CO2 emissions during the period of standing water were 144± 18.18 and 111± 9.51 mg CO2–C m−2 h−1, respectively, for the LFS and SFS sites. During the growing season, when there was no standing water, the CO2 emissions were significantly larger from the LFS (244± 24.88 mg CO2–C m−2 h−1) than the SFS (183± 14.90 mg CO2–C m−2 h−1). Fluxes of N2O ranged from −0.37 to 0.65 mg N2O–N m−2 h−1 at the LFS and from−0.50 to 0.55 mg N2O–N m−2 h−1 at the SFS, with the larger emissions associated with the presence of standing water at the LFS but during the growing season at the SFS. Overall, soil temperature and moisture were identified as the main drivers of the seasonal changes in CO2 fluxes, but neither adequately explained the variations in N2O fluxes. Analysis of total C, N, microbial biomass and Q10 values indicated that the higher CO2 emissions from the LFS were linked to the flooding-associated influx of nutrients and alterations in soil microbial populations. These results demonstrate that annual CO2 and N2O emissions can be higher in longer-term flooded sites that receive significant amounts of nutrients, although this may depend on the restriction of diffusional limitations due to the presence of standing water to periods of the year when the potential for gaseous emissions are low.

fitting another similar 15 cm long tube just before the onset of the flooding period. To prevent the chamber from being displaced by any wind during sampling, four thin rods were fixed into the ground around each sampling point to maintain the chamber in position. Sampling of gases was generally undertaken two to four times each month, but less frequently in the winter months.
Measurements of the CO 2 and N 2 O concentration inside the chamber were made using a Photoacoustic gas 5 analyser (PAS) (Innova 1412, Denmark), connected to the chamber using Teflon tubing. The tubes were 6 m long with a 4 mm inner diameter and the inlet and outlet of the PAS connected to two ports on the top of the chamber. For sampling, the chamber was placed over the collar for between 5 and 6 minutes during which time the gas concentration was analysed 5 to 7 times to complete one sample. Fluxes of CO 2 and N 2 O (mg m -2 hr -1 ) were calculated using: 10 F = (∆C/∆t) (V/A) Where ∆C/∆t : the rate of change in gas concentration inside the chamber during the chamber placement period, which was calculated by fitting a best fit linear regression line to this data versus time; V : chamber volume (4.069 x 10 -3 m 3 ); and A : area bounded by the chamber (0.016 m 2 ). Fluxes of CO 2 and N 2 O were computed if linear regressions produced r 2 > 0.90 (P<0.05) for CO 2 and r 2 > 0.70 (P<0.05) for N 2 O. Annual CO 2 and N 2 O emissions for Feb. 2014-Feb. 2015and May, 2014-April, 2015 were computed by linear interpolation of fluxes for each sampling date. The area under the curve was calculated using the trapezoid rule by integrating the area for 12 month periods. To estimate and compare the contribution of CO 2 and N 2 O fluxes to the Global warming potential (GWP), N 2 O was converted to CO 2 -equivalents by multiplying it by 298 (Solomon et al., 2007).

20
Values of Q 10 for CO 2 emissions were computed for the SFS and LFS using the equation Q 10 = exp (slope*10), where the slope was derived from the regression coefficient of the exponential equation fitted to the CO 2 flux and temperature data.

Environmental measurements
Along with the flux measurements, other environmental variables that could potentially influence the GHG 25 fluxes were also measured. A weather station, located about ~ 100m from the locality where measurements were made, comprised sensors for air temperature (RHT3nl-CA), humidity (RHT3nl-CA), solar radiation (PYRPA-03) and rainfall (RG2+WS-CA). Average air temperature and cumulative rainfall were recorded at 2 m height every 5, and 60 minutes, respectively. Soil moisture content and temperature were measured adjacent to the collars/chambers using a hand-held Theta probe (Delta-T Devices Ltd., Cambridge, UK) each time gas 30 sampling was performed. The depth of standing water (WD) was measured adjacent to the gas sampling points using a graduated wooden ruler. Redox potential was measured from each collar using a portable Hanna redox meter (HI9125, Hanna Instruments) with a 10 cm redox electrode. Redox potential was measured from the soil surface except when the LFS was flooded above a height of 10 cm, in which case the measurements were Biogeosciences Discuss., doi:10.5194/bg-2016-522, 2017 Manuscript under review for journal Biogeosciences Published: 3 January 2017 c Author(s) 2017. CC-BY 3.0 License. acquired from the surface of water. Complete insertion of the electrode to the top soil was avoided to prevent the uncertain impact of the intrusion of water into the electrode through the rim at the top during sampling.

Soil sampling for physical and chemical analysis
Sampling of soil from the two sites (each n = 6) for analysis of its physicochemical properties was carried out in July 2015 and the samples were then air dried and sieved (2 mm) before analysis. Soil texture was measured

Microbial biomass
Microbial biomass C (MBC) and N (MBN) was determined using the chloroform fumigation extraction method (Vance et al., 1987). 10 g samples of fresh soil (n=4-6) from each site were fumigated in a desiccator with 20 ml ethanol-free chloroform for 72 hours and then extracted with 0.5 M K 2 SO4. Identical numbers of subsamples were extracted with the same solution but without fumigation the day after sampling. Supernatants from both 25 fumigated and non-fumigated samples were filtered through Whatman No. 1 filter paper and stored in the freezer until analysis. Organic carbon and total nitrogen in the filtrate were analysed using a TOC/TN analyser (Shimadzu, Japan). Estimates of MBC and MBN were derived by calculating the difference between the results of the corresponding fumigated and non-fumigated analysis, divided by the extraction efficiency factor. Factors of 0.45 (Vance et al., 1987) and0.54 (Brookes et al., 1985) were used for MBC and MBN, respectively, to 30 account for uncompleted extraction of C and N in the microbial cell walls (Jonasson et al., 1996).

Biogeosciences
Beta-glucosidase (BG) was determined using the method described by Eivazi and Tabatabai, (1988). After placing 1 g of soil in a 50 ml flask, 0.25 ml toluene, 4 ml of modified universal buffer (pH 6.0) and 1 ml β-Dglucoside and p-nitrophenyl-α solutions (PNG) were added sequentially and mixed by swirling. Samples were then incubated at 37 0 C for 1 hr, following which 1 ml of 0.5 M CaCl 2 and 4 ml of 0.1 M of tris (hydroxymethyl) aminomethane (pH 12) were added to halt further reactions. The supernatants were filtered and 5 the absorbance of the filtrate measured at 410 nm using a spectrophotometer (Beckman Coulter, DU 530, UV/vis spectrophotometer). For control samples, the same procedure was followed except that PNG was added just before filtering the soil suspension instead of adding it at the beginning.
Protease activity (PRO) was determined as described by Kandeler et al., (1999). After 5 ml of sodium caseinate solution was added to 1 g soil, the samples were incubated at 50 0 C for 2 hours and then filtered after adding 5 10 ml of trichloroacetic acid solution. Alkali and Folin-Ciocalteu's reagents were added to the filtrates before protease activity was determined colorimetrically at 700 nm.
To determine nitrate reductase activity (NR), 4 ml of 2, 4-Dinitrophenol solution and 1 ml of KNO 3 were added to 5 g samples of soil. After incubation at 25 0 C for 24 hours, 10 ml 4 M KCl was added and filtered. To 5 ml of the filtrate, NH 4 Cl buffer (pH 8.5) sulphanilamide reagent was added, and the activity of the enzyme nitrate 15 reductase was measured colorimetrically at 520 nm.
Total microbial activity was assayed using fluorescein diacetate (FDA), based on the method described by Schnürer and Rosswall, (1982), and later modified by Green et al. (2006). Sodium phosphate buffer (pH 7.6) and FDA lipase substrate were added to flasks containing 1 g samples of soil and incubated for 3 hours at 37 0 C.
The fluorescein content in the filtered sample was measured at 490 nm.

Statistical analysis
All statistical analyses were performed using Minitab 16. All the values reported are means of three to six replicates and standard errors were included when required. To investigate the effects of flooding, we tested for significant differences between the two sites (i.e. LFS and SFS) with different hydroperiods, depending on the state of the sites in terms of water-logging (i.e. during and after inundation) over the study period. This was 25 carried out by applying analysis of variance (ANOVA) for each flux, soil enzymatic activity, TC, TN, mineral N and microbial biomass. Functional relationships between potential environmental drivers and the fluxes of CO 2 and N 2 O were performed using linear or exponential regression models. Multiple regression analysis was used to determine the relative contribution of more than one independent environmental driver on CO 2 and N 2 O fluxes. Normality and homogeneity of the variance of all the models were checked visually from residual versus 30 fitted plots and, when necessary, either square-root or log transformations applied. Differences were considered statistically significant when P<0.05, unless otherwise mentioned. Discuss., doi:10.5194/bg-2016Discuss., doi:10.5194/bg- -522, 2017 Manuscript under review for journal Biogeosciences Published: 3 January 2017 c Author(s) 2017. CC-BY 3.0 License. Results

Soil Characteristics
The relative proportion of clay is higher at the LFS, but both sites have sandy loam soil textures. Soil pH was higher at the SFS than the LFS, and increased with depth at both sites (Table 1). Porosity at the LFS was greater than at the SFS. The soil C and N concentrations were significantly higher (P<0.001) at the LFS site, with the 5 greatest difference in the upper soil layers. At both sites, C and N decreased with soil depth, but more gradually at the SFS ( Fig. 2a, b).

Rainfall, water depth, redox potential and air temperature
Depending on the timing and duration of flooding the following periods could be identified, as indicated on associated with longer periods of inundation. In contrast oxidising conditions were always found at the SFS site and these were consistently higher (more positive) than those found at the LFS site (Fig. 3d). Measurements made on the water column, however, indicated that this was always oxidising and had a low but positive redox potential (Fig. 3d).

30
The results are presented with reference to the periods A-E, which represented the state of the sites in relation to water availability. The CO 2 ( positive correlations were also found between CO 2 emissions and soil temperature for both SFS (R 2 = 0.44, P<0.001) and LFS (R 2 = 0.56, P<0.001), with a somewhat greater response at LFS (Fig. 5a). A significant negative linear relationship between soil moisture and CO 2 emission was found at the LFS (R 2 = 0.52, P<0.001) and the SFS (R 2 = 0.54, P<0.001) (Fig. 5b). Correlations between CO 2 emissions and redox potential were low with R 2 values of 0.25 (LFS) and 0.16 (SFS), respectively, but significant at P<0.05. CO 2 emissions at the LFS 5 were exponentially correlated with water depth (R 2 = 0.45, P<0.001) (Fig. 5c). Combinations of soil temperature and soil water content in multiple regression analysis only resulted in a small increase in explanatory power to ~ 58 and 66 % at the SFS and LFS, respectively, but no relative contribution of redox potential was found in this analysis.
No significant relationship was observed between soil N 2 O fluxes and any of, soil temperature, soil moisture, 10 redox potential or water depth at the LFS. At the SFS, soil N 2 O fluxes did correlate positively with soil moisture (R 2 = 0.13, P<0.01) and soil temperature (R 2 = 0.13, P<0.01), but with a low explanatory power. The N 2 O flux was also not correlated with redox potential at the SFS.

Soil enzymatic/microbial activity
While there are some significant differences on different sampling dates for BG, FDA and PRO, overall they 15 were of similar magnitude and showed similar variation at both sites (Fig. 6). For period B, BG activity was significantly lower (P = 0.017) at the LFS, however, in the second flooding period (the first D), BG was significantly higher (P = 0.001) at the LFS than at the SFS. In the later period, E, FDA activity at the LFS was significantly higher (P = 0.001) than at the SFS. In contrast, NR activities were consistently and significantly lower (P < 0.001) at the SFS and independent of water status (standing water availability). 20 3.6 Microbial biomass and soil NO 3 and NH 4 + Seasonal variations in microbial biomass (MB) appear to differ between the LFS and SFS (Fig. 7). Total MBC was generally higher at the LFS than at the SFS at each sampling period, but particularly significant (P < 0.01) for periods B, late D and early E (Fig. 7a). On only two sampling dates was MBN at the LFS slightly lower than at the SFS; for the remaining dates, higher MBN values were observed at the LFS (Fig. 7b). MBC: MBN ratios 25 were significantly higher (P < 0.01) at the LFS during periods of standing water (Fig. 7c).
The concentrations of NH 4 + and NO 3 were generally higher at the LFS than at the SFS (  . 3c).

30
The annual CO 2 emissions were 11.24 -11.5 and 8.18 -8.76 Mg CO 2 -C ha -1 y -1 from the LFS and SFS, respectively. Longer term flooding therefore increased, rather than reduced, the annual emissions by

Relationships between CO 2 fluxes and water and nutrient availability
The overall negative relationship between CO 2 emissions and soil moisture (Fig. 5b) suggests that flooding or 10 high soil water availability through the creation of low oxygen conditions impedes the decomposition processes that lead to CO 2 production. Standing water would also act as an additional constraint on annual emissions by acting as a physical barrier to gaseous diffusion. However, somewhat paradoxically, larger annual CO 2 emissions were asssociated with the site with the longer flooding period. However, the highest CO 2 emissions and the period when the differences in CO 2 emissions between the two sites were greatest occurred in the 15 summer season after the disappearance of standing water, when the soil was better oxygenated (Fig. 3d). No significant differences in CO 2 fluxes between the LFS and SFS were observed during other parts of the year.
Whilst the presence of standing water during the autumn/winter months could constrain CO 2 emisions at the LFS by acting as a gaseous barrier, the similar values found for the SFS for the same period indicates that this is unlikely to have a significant impact on the annual emissions. Reductions in mineralisation caused by low 20 temperatures may be the more significant factor at these times of the year, consistent with the strong correlations between CO 2 emissions and temperature that were observed in this study. Clearly, the specific impact of flooding on annual CO 2 emissions could therefore depend critically on the timing of flooding events.
Higher CO 2 emissions were obtained at the LFS during the drier parts of the year, when there were similar values for soil moisture/soil temperature at both sites. This may be related to higher organic matter content and 25 nutrient status and a generally higher microbial biomass. Compared to soils supplied with no or little organic matter/nutrients, soils that have received more organic matter are likely to emit substantially larger amounts of CO 2 (Winton and Richardson, 2015). The availability of organic matter is considered one of the most important factors controlling the production of GHGs in wetlands (Badiou et al., 2011). This is often derived from plant production but can be introduced by incoming flood water. In our study, both the difference in total C and N 30 values between the LFS and SFS sites and, specifically, the rapid decline in these nutrients down through the soil profile indicate these are derived largely from external sources, rather than from in situ, plant-related material. Had the carbon been contributed mainly from the plant community, similar or higher carbon contents The redox potential of the SFS and LFS sites were, however, different and the ranges at which the highest CO 2 emissions occurred were 220-362 mv and 145-259 mv at the SFS and LFS, respectively. The data could imply 5 that the larger CO 2 fluxes were enhanced by more oxidized soil conditions. However, they were not dependent on the redox level of the soil as the highest CO 2 emissions at the LFS occurred in the lower end of the redox range observed. The lower redox potential at the LFS after the disappearance of standing water could be due to the free-draining nature of this site and/or the presence of a higher organic matter content. Gardiner and James, repectively, as the conditions conducive for N 2 O production.

20
Several studies have shown a negative relationship between GHGs and water level in riparian wetlands (e.g. Mander et al., 2015;Marín-Muñiz et al., 2015). Increases in water depth at the LFS during the flooding period were accompanied by a decrease in the rate of CO 2 emissions, perhaps due to a decrease in near-surface oxygen supply as a result of high standing water levels. Water depth has been shown to be more significant than temperature in determining the variation in CO 2 fluxes during the inundation period (Dixon et al., 2014).

25
Multiple regression analysis of CO 2 fluxes showed a significant paraboloid relationship (R 2 = 0.62, P<0.001) with water depth and soil temperature combined (Fig. 8) but most of this variation is explained by changes in water depth alone (R 2 = 0.45, P<0.001) (Fig. 5c). Peak CO 2 fluxes (above 75 mg CO 2 -C m -2 h -2 ) were recorded during Periods B and D when less than 9 cm water depths were coupled with soil temperatures above 11 0 C. No major variations in CO 2 fluxes were observed when the water depths were greater than 12 cm above the soil 30 surface. Even though some studies have shown a significant relationship between N 2 O fluxes and water depth (Mander et al., 2015;Marín-Muñiz et al., 2015), no correlation was found in this study (Section 3.4). Audet et al., (2013) also found no significant impact of water depth on N 2 O emission from a temperate riparian wetland. However, this may not always be the case and Jacinthe, (2015) showed that some terrestrial riparian ecosystems, which were exposed to different flooding frequencies, routinely acted as a strong sink for CH 4 , except for a 30 small contribution in emissions from permanently flooded sites.

Conclusion
This study provides evidence that the interaction of a grassland ecosystem with the hydrologic regime, impacts on the annual emissions of greenhouse gases. Flooding duration affected the dynamics of CO 2 and, to a lesser