The impacts of drainage, nutrient status and management practice on the full carbon balance of grasslands on organic soils in a maritime temperate zone

Temperate grasslands on organic soils are diverse due to edaphic properties but also to regional management practices and this heterogeneity is reflected in the wide range of greenhouse gas (GHG) flux values reported in the literature. In Ireland, most grasslands on organic soils were drained several decades ago and are managed as extensive pastures with little or no fertilisation. This study describes a 2-year study of the net ecosystem carbon balance (NECB) of two such sites. We determined GHG fluxes and waterborne carbon (C) emissions in a nutrient-rich grassland and compared it with values measured from two nutrient-poor organic soils: a deep-drained and a shallow-drained site. Carbon dioxide (CO2), methane (CH4) and nitrous oxide (N2O) fluxes were determined using the chamber technique, and fluvial C fluxes were estimated by combining drainage water concentrations and flows. The nutrient-rich site was an annual source of CO 2 (233 g C m−2 yr−1), CH4 neutral, and a small source of N2O (0.16 g N2O-N m−2 yr−1). Net ecosystem exchange (NEE) at the shallow-drained nutrient-poor site was −89 and −99 g C m−2 yr−1 in Years 1 and 2 respectively, and NEE at the deep-drained nutrient-poor site was 85 and −26 g C m−2 yr−1 respectively. Low CH4 emissions (1.3 g C m−2 yr−1) were recorded at the shallow-drained nutrient-poor site. Fluvial exports from the nutrient-rich site totalled 69.8 g C m−2 yr−1 with 54 % as dissolved organic C. Waterborne C losses from the nutrient-poor site reflected differences in annual runoff totalling 44 g C m −2 yr−1 in Year 1 and 30.8 g C m−2 yr−1 in Year 2. The NECB of the nutrient-rich grassland was 663 g C m−2 yr−1 with biomass exports being the major component accounting for 53 %. The NECB of the nutrient-poor deep-drained site was less than half of the nutrient-rich site (2-year mean 267 g C m −2 yr−1). Although NEE at the nutrient-poor shallow-drained site was negative in both years, high biomass export meant it was a net C source (2-year mean NECB 103 g C m −2 yr−1). While the impacts of the nutrient and drainage status on NEE, biomass exports and fluvial C losses were confirmed, inter-regional differences in management practice and climate were also significant factors which impacted on the overall NECB of these ecosystems. Contrary to expectation, the NECB of nutrient-poor drained organic soils under grasslands is not necessarily a large C source and this has implications for Ireland’s choice of national GHG inventory reporting methodologies. This study can also aid the development of strategies to deliver reduced emissions tailored to local grassland types.


Introduction
Organic soils are characterised by a high content of partially decomposed organic matter and are an important component of terrestrial carbon (C) storage (Gorham, 1991;Garnett et Published by Copernicus Publications on behalf of the European Geosciences Union.al., 2001).Drained peatlands and peat fires are responsible for almost one-quarter of C emissions from the land use sector with at least 2 Gigatonnes (Gt) of carbon dioxide (CO 2 ) emitted to the atmosphere globally each year (Parish et al., 2008;FAO, 2013a).In Europe, 20 % of all CO 2 emissions between 1990 and 2010 originated from drained organic soils (FAO, 2013b).
In Ireland, organic soils contain an estimated 1-1.5 Gt of C, which represents between 62 and 75 % of the total soil C pool (Tomlinson, 2005;Eaton et al., 2008;Renou-Wilson et al., 2011).However, this C store is under threat as centuries of peatland exploitation (peat extraction, agriculture and forestry) have left only ca. 15 % of peat soils in a natural state (Renou-Wilson et al., 2011).Carbon emissions (CO 2 and methane, CH 4 ) from Irish peat soils and related activities are estimated to account for ca. 3 Mt C per year (Wilson et al., 2013b), equivalent to the emissions reported from the transport sector (Duffy et al., 2013).Therefore, managing organic soils as a means of mitigating greenhouse gas (GHG) emissions from agricultural systems can be an effective strategy for lowering national emissions (Dawson and Smith, 2007;Smith et al., 2007).Of note, a new activity called "rewetting and drainage" of organic soils has been included in the next post-Kyoto Protocol commitment period (2013)(2014)(2015)(2016)(2017).New guidelines, together with default emission factors, have been developed for the 2013 Wetland Supplement of the Intergovernmental Panel on Climate Change (IPCC, 2014).
The use of organic soils for agriculture is a contentious land-use option when considering atmospheric impacts.In Ireland, agriculture is the largest contributor to overall national emissions and accounted for 32.1 % in 2012 (Duffy et al., 2013), which is the greatest proportion among EU Member States (Eurostat, 2013).Grassland is the predominant land use in the temperate zone and covers 60 % of the land area of Ireland (EPA, 2008).The reclamation of raw peat soils or cutover peat soils for grassland was a direct result of population pressures in the 19th century which intensified in the 20th century due to national drainage acts and agricultural schemes (Feehan et al., 2008).Presently, some 300 000 ha of organic soils is under grassland (with negligible area cultivated for crops), on par with organic soils drained for forestry and, together with industrial cutaway peatlands, managed peatlands represent nearly half of the total peat soil area or 10 % of the total land area (Wilson et al., 2013b).
Drainage and grassland management practices (e.g.fertilisation) of organic soils stimulate the aerobic oxidation of previously accumulated organic matter, promoting CO 2 and nitrous oxide (N 2 O) emissions while reducing methanogenesis (Byrne et al., 2004;Freibauer et al., 2004;Nieveen et al., 2005;Jacobs et al., 2007;Elsgaard et al., 2012;Drösler et al., 2013;Schrier-Uijl et al., 2013).However, inter-regional spatio-temporal variability of CO 2 flux is high, particularly in western Europe.The implication is that physical and bio-geochemical soil conditions (nutrient content, organic matter quality and water table levels) as well as local management practices and subtle climatic variations exert strong influences on C loss (Drösler et al., 2008).In addition, drainage of temperate organic soils typically leads to higher dissolved organic C (DOC) fluxes (Wallage et al., 2006;Urbanová et al., 2011), so that management practices may alter both contemporary photosynthetic uptake rates, but also the stability of older soil stores of sequestered C. Further, although fluvial C exports are increasingly recognised as potentially significant components of terrestrial C balances, full accounting of all fluvial C forms remains the exception rather than the rule, notably for dissolved inorganic and particulate organic C.
The annual GHG emissions from organic soils under grassland are currently evaluated in Ireland using the generic Tier 1 reporting method (IPCC, 2006).These emission factors typically rely on assessing and averaging available data and so can be biased towards locations with high research activities, and therefore individual regional climatic and edaphic conditions where such research is conducted.For example, grasslands on organic soils in Germany, where much research on grassland GHG balances is conducted, are typically fertilised, deep-drained and nutrient-rich (Drösler et al., 2008(Drösler et al., , 2013)), whereas in Ireland, extensively grazed permanent pasture with partial drainage is common.The primary objective of this study is to estimate annual net CO 2 , CH 4 and N 2 O fluxes as well as fluvial C fluxes from two typical Irish grasslands over organic soil.We further combine estimates of terrestrial net ecosystem exchange (NEE), fluvial C export, and CH 4 fluxes associated with grazing livestock, as well as biomass C export, to calculate the net ecosystem C balance (NECB).By comparing two grasslands with contrasting drainage, soil nutrient status and management practices, we assess the role of these attributes in respect to biomass productivity, GHG and waterborne C fluxes, and appraise the hypothesis that specific site characteristics (low soil fertility and higher water table levels), management systems (low stocking density and low inputs) and regional climate can affect the overall C balance of these ecosystems.Finally, this study provides data to support a progression towards the IPCC Tier 2 reporting level in Ireland by producing emission factors (EFs) for CO 2 , CH 4 , N 2 O and DOC for typical organic soils under grassland.

Study sites
The study sites were located in a maritime temperate climate zone in two distinct Irish ecoregions, both characterised as landscapes rich in organic soils.Site A at Glenvar, Co. Donegal (latitude: 55 • 9 N, longitude: 7 • 34 W) is situated at 40 m elevation in the northwest of the country, less than 1 km from the sea shore, and the climate is typical of western maritime  Mills, 2000).Land use in both regions is predominantly grassland and the two sites are managed for low intensity beef suckler production.Site A has a lower stocking density of 0.6 Livestock Units (L.U.) per ha (all suckler cows) compared to Site B with 1.2 L.U. (sheep in the winter months and suckler cows and calves in the summer months).While Site A only receives on-site organic fertilisation from manure directly deposited by cattle in the field, artificial fertiliser is usually applied in spring to Site B at a rate of 25 kg N ha −1 .However, no fertilisation took place at Site B during the first monitoring year.No cows are present on-site during the winter months and no concentrates or additional feeding occur on site.
Both Sites A and B have been drained for over 60 years and the soils are categorised as terric and limnic histosols respectively (FAO, 1998).With an organic matter content higher than 30 % (Table 2) over a depth greater than 40 cm, they satisfy the definition of peat soils in both the Irish soil classification (Hammond, 1981) and the FAO key to soil types (FAO, 1998), which form the basis of definitions used in the IPCC guidelines (IPCC, 2006(IPCC, , 2014)).They are typical "earthy peat", i.e. reclaimed drained peat soils characterised by a well aerated structure and a relatively firm surface horizon containing few or no recognisable plant remains.The peat in Site A overlies bedrock of Precambrian schist and gneiss and contains on average 20 % C (Table 2).The base geology of Site B is of limestone origin and the overlying peat has a higher pH (5.5) and almost double the C content (40 %) and is situated over a thick layer of marl (pH 7.2).Despite similar C : N ratios (16-21), Sites A and B differ in nutrient status and, based on N, phosphorus (P) and potassium (K) content, Site A is nutrient-poor while Site B is nutrientrich (Table 2).
Site B has a uniform hydrological soil profile throughout the experimental area.In contrast, Site A can be divided into two distinct sub-sites: a deep-drained area "Site A d " (defined in the 2006 IPCC guidelines as the mean annual water table depth of 30 cm and deeper below the surface) and a shallowdrained area "Site A s " (defined as the mean annual water table depth of less than 30 cm below the surface).

Vegetation
Both sites displayed a moderate to low quality sward dominated by Holcus lanatus and Agrostis spp.Wetland species, such as Equisetum palustre, Juncus effusus and Juncus articulatus were present in the shallow-drained Site A s while Site B was characterised by a greater cover of Rumex acetosa and Cirsium arvensis.While the moss layer was almost non-existent in Site B, Site A d and A s had an average ground cover of 20 and 60 % respectively, composed solely of Rhytidiadelphus squarrosus.Experimental plots were fenced off from cattle for the duration of the study.To simulate local activities, the vegetation within and surrounding each collar for GHG measurements was cut to a stubble height of 5 cm in order to mimic the cow grazing regime, so that the height of the vegetation was always comparable to that in the grazed fields.The interval between cuttings (considered as grazing events) was not strict but rather followed the concomitant status of the vegetation in the fields.The number of cuttings also varied each year reflecting the biomass production.The vegetation was collected from each plot and oven dried to a constant mass at 75 • C. C and N contents were measured (CE440 Exeter Elemental Analyser) using a pooled, homogenised sample of all biomass cut from each plot from the first growing season.Vegetation height (cm) was measured regularly throughout the year and systematically during GHG measurements and before cutting events.Conjunctly, leaf area index (LAI) was measured at each plot when photosynthetic photon flux density (PPFD) levels were sufficient (November-March excluded).The Accu-PAR LP-80 (Ceptometer, Decagon Instruments, WA, USA) was used to measure light interception via 80 independent photo sensors.The PPFD transmission data were used to calculate gap fractions which were inverted to derive LAI estimates (Norman and Campbell, 1989).Polynomial regression constrained to pass through the origin was used to develop site-specific relationships between vegetation height and LAI (Fig. 1) for use in Eq. (1) (see below).

Greenhouse gas sampling
Twelve permanent sample plots were established systematically within Site A (n = 7 for Site A d , n = 5 for Site A s ) and nine sample plots at Site B (at least 15 m from field border or river).Each sample plot consisted of a stainless steel collar (60 × 60 cm) that was inserted to a depth of 20 cm into the soil prior to the start of the study.Perforated PVC pipes (internal diameter: 2 cm) were inserted adjacent to each sample plot to measure water table position (WT).Wooden boardwalks were built around the sample plots to minimise damage to the vegetation and to avoid compression of the peat during gas sampling.Data loggers (Hobo External Data Loggers, Onset Computer Corporation, MA, USA) were established at each study site and recorded hourly soil temperatures ( • C) at 5, 10 and 20 cm depths.A weather station (Watch Dog Model 2400, Spectrum Technologies Inc., IL, USA) was established at each site and recorded PPFD (µmol m −2 s −1 ), soil temperatures ( • C) at 5 and 10 cm depths, hourly rainfall (mm) and volumetric soil moisture at 5 cm (%, Site B only).Site A was monitored starting April 2011 (first 2 years are reported here) while Site B was only monitored for 1 year (April 2011 to March 2012) with no fertiliser application taking place during that time.Additional N 2 O measurements were taken at Site B in April-June 2012 during an intensive campaign after a typical fertilisation event of 25 kg N ha −1 .

CO 2 flux measurements
CO 2 fluxes were measured from April 2011 to March 2013 (Site A) and from April 2011 to March 2012 (Site B) at fortnightly (summer) and monthly (winter) intervals using the static chamber method (Alm et al., 2007).Measurements were done between 8 a.m. and 6 p.m. (2 to 9 measurements per collar per day).Instantaneous NEE was measured over a range of PPFD using a transparent polycarbonate chamber (60 × 60 × 33 cm) equipped with internal fans to ensure mixing of the air and a cooling system to maintain the temperature within the chamber close to the ambient air temperature (see Alm et al., 2007).For each measurement, the chamber was placed in a water-filled channel at the top of the collar and CO 2 concentration (ppmv) in the chamber headspace was measured at 15 s intervals over a period of 60-180 s using a portable CO 2 analyser (EGM-4) (PP Systems, UK).PPFD was measured by a quantum sensor located at the top of the chamber (PAR-1.PP Systems).Concurrently, air temperature ( • C) within the chamber and soil temperatures at 5, 10 and 20 cm depths were recorded at each collar (soil temperature probe; ELE International, UK).The WT position relative to the soil surface was manually measured with a water level probe (Eijkelkamp Agrisearch Equipment, the Netherlands).Following each NEE measurement, the chamber was vented for a short time by removing it from the collar.This was carried out in order to ensure equilibration of the gas concentration.The chamber was then replaced in the collar and covered with an opaque material in order to provide an estimate of ecosystem respiration (R eco ), which is the sum of heterotrophic and autotrophic respiration enclosed by the chamber.Flux rates (mg CO 2 m −2 h −1 ) were calculated as the linear slope of the CO 2 concentration in the chamber headspace over time, with respect to the chamber volume, collar area and air temperature.A flux was accepted if the coefficient of determination (r 2 ) was at least 0.90.An exception was made in cases where the flux was close to zero (mainly in winter time where soil/plant processes are typically slower) and the r 2 is always low (Alm et al., 2007;Görres et al., 2014).In these cases (∼ 2 % of total fluxes) the flux data were examined graphically and fluxes with obvious non-linearity (due to chamber leakage, fan malfunction etc.) were discarded.The remainder were accepted provided that some of the environmental variables measured at the same time (e.g.soil temperature, PPFD, LAI) were sufficiently low to account for the low flux values (e.g Wilson et al., 2013a).Gross primary production (GPP) was calculated as the sum of NEE and R eco values (Alm et al., 2007).

CH 4 and N 2 O flux measurements
CH 4 and N 2 O fluxes were measured at fortnightly/monthly intervals using the static chamber method, which consisted of an opaque, polycarbonate chamber (60 × 60 × 25 cm) equipped with a battery-operated fan, which mixed the air within the chamber headspace.Four 50 mL samples were withdrawn into 60 mL polypropylene syringes from the chamber headspace at 10 min intervals over a 40 min period (first sample taken 10 min after chamber closure) and then injected into pre-evacuated Exetainers ® vials (12 mL Soda Glass Vials, Labco, UK).During each measurement, air temperature inside the chamber, soil temperature (at 5, 10 and 20 cm depths) and WT were recorded at each collar.Gas samples were sent to Justus-Liebig Universität Gießen, Germany for analysis.The samples were analysed for CO 2 , CH 4 and N 2 O concentrations with a gas chromatograph (Bruker Greenhouse Gas Analyzer 450-GC) equipped with a thermal conductivity detector (TCD), a flame ionisation detector (FID) and an electron capture detector (ECD).Detector temperatures were 200 • C (TCD), 300 • C (FID) and 300 • C (ECD).The CO 2 (304,402,1509,4000 and 20 005 ppm), CH 4 (1.02, 1.81, 5.02, 20.9 and 101.9 ppm) and N 2 O standards (0.25, 0.32, 2.01.15.1 and 100.1 ppm) standards were supplied by Deuste Steininger GmbH.Gas peaks were integrated using Galaxie software (Varian Inc. 2006).Fluxes (mg CH 4 m −2 h −1 and µg N 2 O m −2 h −1 ) were calculated from the linear change in gas concentration as a function of time, chamber volume, collar area and air temperature.A flux was accepted if the coefficient of determination (r 2 ) was at least 0.90.The same criteria applied to CO 2 in regard to low fluxes were also applied to CH 4 and N 2 O fluxes.Approximately 32 % of CH 4 fluxes and 15 % of N 2 O fluxes were discarded for obvious non-linear changes over the measurement period.Positive flux values indicated a loss of CH 4 and N 2 O to the atmosphere and negative flux values indicated CH 4 and N 2 O uptake.

CO 2 flux modelling
Statistical and physiological response models (Alm et al., 2007) were constructed and parameterised for each study site.Model evaluation was based on the following criteria; (a) statistically significant model parameters (p < 0.05), (b) lowest possible standard error of the model parameters and (c) highest possible coefficient of determination (adjusted r 2 ) (see Laine et al., 2009).Given the homogeneous dominant vegetation within each study site, we combined data from all sample plots (n = 12 at Site A and n = 9 at Site B) to construct a single model for gross photosynthesis (GPP) and ecosystem respiration (R eco ) for each site.Model coefficients were estimated using the Levenberg-Marquardt multiple non-linear regression technique (IBM SPSS Statistics for Windows, Version 21.0.Armonk, NY, USA).One-third of the data was randomly removed from all data sets prior to www.biogeosciences.net/11/4361/2014/Biogeosciences, 11, 4361-4379, 2014 modelling and used to independently test the models.During model construction, the relationship between GPP or R eco and a range of independent environmental variables (recorded in conjunction with flux measurements) was tested.
Only variables that increased the explanatory power of the model were included.The models were accepted if the residuals were evenly scattered around zero.In the GPP models, we used PPFD and LAI as explanatory variables: where P max is maximum photosynthetic rates, PPFD is photosynthetic photon flux density, k PPFD is the PPFD value at which GPP reaches half its maximum (half saturation constant), LAI is Leaf Area Index (see Sect. 2.2) and a is a model parameter.
In the R eco models, we used soil temperature at 5 cm depth (T 5 cm ) and either WT level (Eq.2) for Site A or volumetric moisture content (VMC) (Eq. 3) for Site B: where R eco is ecosystem respiration, T REF is reference temperature set at 283.15 K, parameter T 0 is the (minimum) temperature at which respiration reaches zero and set here at 227.13 K, VMC is volumetric moisture content, WT is water table depth and a, b and c are model parameters.

Reconstruction of annual CO 2 -C balance
The response functions estimated for GPP and R eco were used for the annual reconstruction of NEE.In combination with an hourly time series of (1) PPFD, VMC (Site B) and T 5 cm , recorded by the weather station and data loggers, (2) modelled LAI and (3) WT depths linearly interpolated from weekly measurements, GPP and R eco fluxes were reconstructed for each sample plot.NEE was then calculated on an hourly basis as follows: NEE = GPP -R eco (Alm et al., 1997).Negative NEE values indicated a net uptake of CO 2 from the atmosphere by the peatland and positive values indicated a net loss of CO 2 to the atmosphere.The annual CO 2 -C balance (g C m −2 yr −1 ) was calculated for each sample plot by integrating the hourly NEE values over each 12 month period (Year 1: 1 April 2011 to 31 March 2012 and Year 2: 1 April 2012 to 31 March 2013).Uncertainty in the reconstructed annual GPP and R eco was calculated by summing up the maximum and minimum standard errors associated with each of the model parameters (e.g.Drösler, 2005;Elsgaard et al., 2012).As NEE is not directly modelled, uncertainty in the annual NEE estimate was calculated following the law of error propagation as the square root of the sum of the squared standard errors of GPP and R eco .Modelled CO 2 fluxes, GPP and R eco , as well environmental parameters were tested using repeated measures GLM (General Linear Models) to show any effects of sites and year and combination of site and year (IBM SPSS Statistics for Windows, Version 21.0.Armonk, NY.USA).

Waterborne C exports
Fluxes of DOC, particulate organic C (POC), and dissolved inorganic C (DIC) were determined for each site for the hydrological year beginning October 2011.For Site A, a second year of monitoring covered the hydrological year commencing October 2012.Discharge measurement, laboratory analyses and C flux estimation methods employed are briefly described below and detailed in Barry et al. (2014).
In order to accurately estimate waterborne C fluxes representative of the area on which the terrestrial GHG fluxes were conducted, runoff measurements were confined to the smallest catchment areas encompassing the GHG study sites.Field drains adjacent to each of the study sites were selected and these drained relatively small areas of between 3 and 5 ha.Critically, soils and land use within these catchments was uniform and consistent with that of the GHG study sites.Although areas outside of the study sites yielded water, this approach has advantages in that sampling is close to source, limiting the residence time between exports from the soil environment and thereby lessening the impact of processes that can alter aqueous fluxes (e.g.remineralisation and atmospheric evasion of CO 2 ).Continuous level recorders (Orpheus mini, OTT Germany) and flow gauging instrumentation were installed at each site.C concentrations were determined at fortnightly to monthly intervals throughout each monitoring period.DOC concentration was measured by high temperature Pt-catalysed combustion on glass fibre filtered samples (Whatman GF/C; 1.7 µm), POC concentrations by loss on ignition of samples filtered onto glass fibre filters, and DIC was calculated according to Stumm and Morgan (1996), from alkalinity (gran titration), and in situ measurements of pH, temperature and specific conductance.Annual C loads were calculated as the product of annual or seasonal flow-weighted mean concentrations and runoff volumes, corresponding to "method 5" given by Walling and Webb (1985), and DOC rainwater inputs were subtracted based on concentrations determined for rainwater during the study.Fluxes are expressed as areal rates according to catchment area.95 % confidence limits for flux estimates were calculated according to Hope et al. (1997).
As DIC reflects both CO 2 derived from soil respiration and C derived from the atmosphere and rock weathering, total DIC exports were apportioned to reflect terrestrial organic matter dynamics alone (Stumm and Morgan, 1996;Telmer and Veizer, 1999).All CO 2 above atmospheric equilibrium was considered derived from respiration of organic matter.
The origin of C derived from rock weathering present as dissolved carbonates varies with respect to the C content of the source geology, such that 50 % of the C present as bicarbonate from the dissolution of carbonate rocks originates from the rock, and 50 % from carbonic acid (with soil respiratory/atmospheric origin).In contrast, all C present as bicarbonate from the dissolution of feldspars and silicate rocks originates from carbonic acid.Thus for Site B, which overlies limestone geology, half of the bicarbonate export was attributed to soil respiratory C, whereas for Site A, which overlies Precambrian Gneiss and Schist, all bicarbonate was attributed to soil respiratory C.

Calculation of NECB
A full net ecosystem carbon balance (NECB) can be derived from the terrestrial GHG (CO 2 and CH 4 ) sources and sinks by adding fluvial C fluxes and C emissions associated with biomass removal and the livestock grazing (enteric fermentation and on-site manure deposits).The following nationally derived emission rates for CH 4 from enteric fermentation were employed as detailed by O'Mara et al. ( 2007): 74 kg CH 4 head −1 yr −1 for suckler cows; 22.4 kg CH 4 head −1 yr − for calves; 8 kg CH 4 head −1 yr −1 for sheep.

Weather conditions
In Year 1, Site A displayed higher precipitation (1211 mm) as well as a higher annual mean air temperature (10.2 • C) compared to Site B (1003 mm; 9.6 • C) (Table 1).Rainfall and temperature patterns were similar with both sites having higher mean annual precipitation (7-12 %) and air temperature (5-10 %) compared to the 30-year average values (Table 1).Both sites also had cooler summer but warmer autumn and winter temperatures compared to the 30-year monthly averages.There was no significant difference between monthly mean soil temperatures between all the sites (p = 0.801) (Fig. 2).In Year 2, Site A again received higher precipitation that the long-term average (1193 mm) driven by high values during June and July, September and October, and December and January.However, both the mean annual and monthly air temperatures were consistently below the long-term averages (by 8-9 %).
There was a typical seasonal trend in mean monthly PPFD values at both locations (Fig. 2) associated with day length and seasonal cloud cover, with the highest values observed in summer (June to August) and lowest in the winter (December to February).Annual and mean monthly PPFD values were significantly higher (p < 0.001) at Site B than Site A during Year 1.There was no significant difference between Year 1 and Year 2 at Site A but PPFD values were consistently higher in Year 2 during the period June-December, except for July.Seasonality of precipitation was reflected in WT depth at Site A and soil moisture content at Site B (Fig. 3).Despite rising slightly above −40 cm in winter, the monthly WT site averages at Site A d were relatively stable and similar annual means were recorded in both years (−47.8 and −47.4 cm).Monthly means were significantly different between Sites A d and A s (p < 0.001), the latter being characterised by seasonal fluctuations with monthly WT below −30 cm during late summer/early autumn periods, rising quickly to levels above −10 cm during winter periods.Despite this, the annual WT mean at Site A s was similar between years at −22.9 cm in Year 1 and −24.2 cm in Year 2. At Site B, VMC averaged 47.5 % and ranged from 20 to 62 % with two significantly drier periods in May and August in Year 1.

CO 2 fluxes
The relationships between GPP and the environmental variables differed between the study sites as demonstrated by the different model coefficients derived for Eqs. ( 1)-(3), and given in Table 3. GPP is strongly dependent on irradiance (PPFD) and is commonly described by a rectangular hyperbola.Seasonal variation in the photosynthetic capacity of the vegetation, described by LAI, was incorporated into the model in a manner similar to that described by Wilson et al. (2007b).A close relationship between GPP and PPFD was observed at all sites (r 2 = 0.53-0. of the LAI term further improved the explanatory power of the model for both Site A (r 2 = 0.72) and Site B (r 2 = 0.85).Validation of the model showed a strong agreement between modelled versus observed GPP fluxes (Fig. 4) and independent test data (r 2 = 0.86 at both Sites A and B).However, higher variation was evident in the predicted GPP fluxes at Site B, particularly at low to medium flux rates (1000-3000 mg CO 2 m −2 h −1 ).At Site A, R eco was driven by T 5 cm and WT (r 2 = 0.75) and at Site B, R eco was controlled by VMC and T 5 cm , and this combination explained 67 % of the variation.As with GPP, the relationship between observed and modelled R eco was generally strong (Fig. 4).At Site B, there was a tendency for the model to overestimate low flux rates and underestimate higher flux rates (r 2 = 0.63 at Site A and r 2 = 0.54 at Site B using independent test data).As expected, monthly estimates of GPP demonstrated clear seasonal variation with maximum rates occurring during the month of July regardless of sites and year.In Year 1, monthly GPP values were consistently higher in Site B (Fig. 5) than Site A. This was also reflected in the higher P max parameter values in the GPP model (Table 3).Within Site A, monthly GPP was higher at Site A d than Site A s in both years.Monthly R eco followed the same seasonal trend as GPP, which exceeded R eco between April and August in Year 1 at Site A s and similarly at Site A d except for the month of July.NEE was positive (i.e.CO 2 -C source) throughout the winter periods as a result of a greater reduction of GPP compared to R eco .Monthly CO 2 emissions (positive NEE values) were much lower in Year 2 for 4 months (October-January) due to lower winter temperatures (Figs. 2 and 5).
In contrast R eco exceeded GPP for 8 months of the year at Site B, with CO 2 uptake only occurring in May through July.However, the highest monthly CO 2 uptake for any plots (−128 g C m −2 month −1 ) were observed during these months (Fig. 5).
In Year 1, annual GPP was significantly greater (p < 0.001) at Site B (2089 g CO 2 -C m −2 yr −1 ) compared to Sites A d (1651 g CO 2 -C m −2 yr −1 ) and A s (1349 g CO 2 -C m −2 yr −1 ) with the latter being 35 % lower than Site B (Table 4).In Year 2, annual GPP was 10 % lower at Site A s and 5 % lower at Site A d presumably in response to lower temperatures and irradiance in the spring and autumn.In Year 1, annual R eco at Site B (2322 g C m −2 yr −1 ) was almost twice that at Site A s (1260 g C m −2 yr −1 ) with Site A d intermediate (1736 g C m −2 yr −1 ).In Year 2, similar trends to GPP were observed with a decrease in annual R eco of 12 % at both Sites A d and A s (Table 4).
Annual NEE varied between sites and years (Table 4).In Year 1, Site A d was a CO 2 -C source of 85 g CO 2 -C m −2 yr −1 and Site A s was a net sink (−89 g CO 2 -C m −2 yr −1 ).However, both sites were CO 2 -C sinks in Year 2 (−26 and −99 g CO 2 -C m −2 yr −1 respectively).In contrast, Site B was a CO 2 -C source of 233 g CO 2 -C m −2 yr −1 .The site effect was highly significant (p < 0.001) between all sites in Year 1. Similar analysis performed between Site A d and A s for both years generated a significant site × year interaction    (p < 0.001) demonstrating that CO 2 fluxes were significantly different between the sites and between years in Site A.

Biomass production/export
Contrasting annual GPP values between sites and between years are associated with biomass productivity.In Year 1, the cumulative production was superior at Site B (351 g C m −2 ), with Site A d and A s 35 % and 50 % less productive respectively (Table 4).Biomass production in Site B started to increase sooner than Site A and was consistently higher, peaking during the warmest months (July-August) (data not shown).Biomass production also lasted for a longer pe- riod at Site B allowing for biomass removal to take place as late as October (eight cutting events in total).Productivity was lower throughout Year 2 with poor growth during late summer months a particular feature.Lower temperatures throughout the second growing season was likely the driver behind the decrease in productivity with biomass production 18 % lower at Site A d (from 225 to 185 g C m −2 ) and 21 % at Site A s (from 174 to 137 g C m −2 ) (Table 4).Since vegetation density (stems m −2 ) was similar between the shallow- and deep-drained plots, an analysis of growth rates demonstrated that for both years, mean growth rates (over periods between grazing events) were significantly lower in Site A s compared to A d (p < 0.001), except for the first two months of the study (Fig. 7).The contrasting productivity between all sites was also reflected in LAI values, which peaked before cutting events during the summer with higher values at ∼ 2 m 2 m −2 for Site B compared to ∼ 1.5 m 2 m −2 for Site A d and ∼ 0.8 m 2 m −2 for A s (data not shown).In Year 2, maximum LAI at Site A d was also consistently higher than Site A s but the difference was reduced in comparison to Year 1.A significant inverse relationship was found between mean annual water table and mean LAI (growing season) during both years (Fig. 6).
The nitrogen content of the vegetation was similar at both Site A d and A s (2 %) but higher at Site B (3.3 %).Nitrogen export in biomass followed C export trends with the nutrient-rich Site B losing the highest amount at 265 (27) kg N ha −1 yr −1 compared to 107 (45) and 80 (12) kg N ha −1 yr −1 in Site A d and Site A s (data are mean ± standard deviation with n = 7 for Site A d and n = 5 for Site A s ).Compared to Year 1, N biomass exports were 17 to 19 % lower in Year 2 in Site A d and A s respectively.

Methane fluxes
Fluxes of CH 4 were not detectable at the deep-drained sites, namely Site A d and Site B, and were relatively low at Site A s (Fig. 8).A strong seasonal effect could be observed in both years at Site A s (Fig. 8).CH 4 fluxes followed a general trend where the highest values were observed in late summer and the lowest in winter time.However, high flux values for this site (0.9 and 1.4 mg CH 4 m −2 h −1 ) were also observed in April and November.Spatial variation in emissions at Site A s was very evident (Fig. 8) with the higher emissions associated with plots that contained Juncus effusus.No statistical relationship was observed between CH 4 fluxes and any of the environmental parameters.Therefore, in order to calculate an annual CH 4 -C balance, linear interpolation between the observed fluxes was performed and the values integrated over a 12-month period (Beetz et al., 2013).Annual CH 4 -C emissions were not significantly different between years with 1.3 (1.09) g CH 4 -C m −2 yr −1 in Year 1 and 1.4 (1.1) g CH 4 -C m −2 yr −1 in Year 2 (Table 4).

Nitrous oxide fluxes
Fluxes of N 2 O were not distinguishable from zero at Site A during the monitoring period.At Site B, a seasonal trend was observable during Year 1 with the highest values (153 µg N 2 O m −2 h −1 ) observed in October (Fig. 9).N 2 O uptake (∼ 14 µg N 2 O m −2 h −1 ) was observed during May.Following a fertilisation event at Site B in Year 2, N 2 O emissions increased considerably and reached an observed maximum of 484 µg N 2 O m −2 h −1 .As with CH 4 fluxes, no statistical relationship was observed between N 2 O fluxes and any of the environmental parameters.Instead, linear interpolation between the observed fluxes was performed and the values integrated over a 12-month period (note that the integration period was from 1 April 2011 to 31 March 2012 and, therefore, did not cover the fertilisation experiment).Annual N 2 O-N emissions for this period were estimated at 0.16 ± 0.34 g N 2 O-N m −2 yr −1 .

Waterborne C losses
Fluvial exports were measured at the field scale at Site A and so encompass both Sites A d and A s .In Years 1 and 2 the fluvial fluxes totalled 44 g C m −2 yr −1 and 30.8 g C m −2 yr −1 respectively.DOC export rates were 18.1 g C m −2 yr −1 in Year 1 and 4.9 g C m −2 yr −1 in Year 2, representing 41 and 16 % of the total C flux in each year respectively (Table 4).The mean drainage water pCO 2 in Years 1 and 2 respectively was 102 and 146 times in excess of atmospheric equilibrium concentrations and consequently the export of excess CO 2 (above equilibrium) was substantial at 13.4 and 18.5 g C m −2 yr −1 respectively.Excess CO 2 represented 30 and 60 % of the waterborne C flux each year and dominated the export of DIC (excess CO 2 + HCO − 3 ) which overall comprised 44 and 78 % of the total fluvial C flux in each year.As with DOC, POC fluxes were also lower in Year 2, declining from 6.4 g C m −2 yr −1 to 2.0 g C m −2 yr −1 .
The fluvial C flux was greater at Site B at 68.9 g C m −2 yr −1 , comprising 54 % DOC, 35 % DIC and 11 % POC.By comparison to Site A, this waterborne C flux largely reflected a greater export of DOC, particularly relative to the second year at Site A. Despite differences in base geology between sites, the flux of DIC (estimated as derived from soil respiratory processes) was comparable to both years at Site A, but differed in that it was primarily in the form of bicarbonate (29 % of total flux).The implication is that soil respiratory CO 2 at this site yielded greater carbonate dissolution rather than accumulating in soil pore waters to high pCO 2 .

NECB
Combining the nationally derived CH 4 emission rates from enteric fermentation with contemporary stocking rates at the sites yielded further emissions of 3.6 g C m −2 yr −1 for Site A and 9.5 g C m −2 yr −1 for Site B. Given the very low intensity farming systems at both sites, we assume the balance of C import/export from on-site manure deposits to be negligible (there was no fertilisation at Site B in the first year).The NECB of the shallow-drained nutrient-poor grassland site was a small source in both years (134 (75) and 73 (56) g C m −2 yr −1 ) while the deepdrained nutrient-poor site was double with 358 (111) and 193 (64) g C m −2 yr −1 .The NECB of the nutrient-rich drained site was 663 (42) g C m −2 yr −1 (Table 4).

Discussion
The large amount of C stored in organic soils means that careful management of these soils should be a priority to reduce C losses.This should start with more accurate estimates of GHG fluxes and waterborne C losses that reflect regional variability (climatic and edaphic) but also local management practices.In this study, we have investigated contrasting permanent grasslands over organic soils in a maritime temperate climate to evaluate site effects and management practices on the GHG fluxes and waterborne C losses.

Contrasting NEE and driving parameters
Annual NEE differed significantly among sites, with the nutrient-poor shallow-drained site (A s ) a net annual sink of CO 2 -C in both years of the study and the nutrient-poor deepdrained site (A d ) a net sink of CO 2 -C in the second year.In contrast, the nutrient-rich drained Site B was a net annual source of CO 2 -C.Given that there was a tendency for the R eco model to underestimate the higher fluxes at this site (Fig. 4), this NEE value may be a conservative estimate as it is at the lower range of data reviewed from temperate and boreal studies (e.g.Maljanen et al., 2010;Elsgaard et al., 2012;Leiber-Sauheitl et al., 2013).Annual losses of CO 2 from our nutrient-rich drained Site B were more in line with those determined for low-intensity Dutch permanent grasslands on organic soils (Jacobs et al., 2007;Veenendaal et al., 2007) and for extensive wet nutrient-poor grasslands in Germany (Drösler et al., 2013).This corroborates the fact that Site B is a low-intensity grassland (no cultivation and low inputs).The importance of a "grassland type" appears even more apparent in the case of nutrient-poor sites as investigated in this study.Net annual uptake of CO 2 -C in both the deep-drained and shallow-drained nutrient-poor sites under investigation here contrast significantly with studies from similar site types in www.biogeosciences.net/11/4361/2014/Biogeosciences, 11, 4361-4379, 2014 temperate climatic regions where a net annual loss of CO 2 -C has been reported (Drösler et al., 2013;Leiber-Sauheitl et al., 2013).The NEE values for the nutrient-poor shallowdrained sites in this study represent similar sinks (small source in some years) to those reported by Skiba et al. (2013) for grazed drained nutrient-poor organic soils (moorland) in a maritime temperate climate.Our results would therefore point to the significance of edaphic conditions as well as the low intensity management system encountered at our sites.
An indication of such influences can be found in the contrasting GPP and R eco parameters found at all sites and which are discussed below.

Productivity drivers
The nutrient-rich grassland was the most productive with an annual GPP value in the range of those reported from Irish grasslands (Byrne et al., 2005;Gilmanov et al., 2007) and comparable to Danish grassland over organic soils (2100 to 2500 g C m −2 yr −1 ) (Elsgaard et al., 2012) and to other maritime intensive grasslands over peat (2239-2403 g C m −2 yr −1 ) in north Germany (Beetz et al., 2013).Despite GPP being significantly lower in the nutrient-poor sites, the values were remarkably high and more productive than similar grasslands over peat in Germany (Beetz et al., 2013;Drösler et al., 2013;Leiber-Sauheitl et al., 2013).This may be a reflection of the longer growing season due to mild winter temperatures experienced in a maritime temperate climate like Ireland (Wilson et al., 2007b) which is influenced by the North Atlantic Drift.Increased soil temperatures over the last three decades also indicate that the growing season is getting longer in Ireland (Hurtado-Uria et al., 2014).Moreover, this temperate humid climate is not associated with very high temperatures that limit productivity during the growing season via water shortage as experienced in other temperate regions.GPP was closely linked to PPFD at all studied sites and the significantly higher PPFD received by the plants at Site B (located further south than Site A) may explain some of the higher productivity there, especially given that irradiance was highest at the end of the growing season when it is one of the most dominant growth-limiting factors as soil temperature is still high (Fig. 2).However, it cannot explain the contrasting GPP between Sites A d and A s where plants received the same PPFD.Gross photosynthesis depends on vegetation status, commonly estimated by proxies, such as LAI, which captures well the variations in vegetation growth that occur both seasonally and as a result of grazing events (e.g.Fig. 1).Our CO 2 measurement campaigns covered the full range of LAI and vegetation heights, which were always highest in the nutrient-rich site.In comparing the two nutrient-poor sites, a significant inverse relationship was found between mean annual water table and mean LAI (growing season) during both years (Fig. 6).Higher water table levels most likely reduced biomass productivity by slowing the rate of gas diffusion through the vegetation (Lo-hila, 2008), effectively inhibiting photosynthesis (decreased GPP).The lower growth rates observed in Site A s could also be explained by the presence of wetland plant species in some of the plots, such as Juncus spp.and Equisetum spp.which have cylindrical leaves and thus affect leaf area estimations (Chen and Black, 1992;Wilson et al., 2007a).As such, local climate and the nutrient and drainage status have been identified as the most important factors explaining the contrasting biomass productivity and consequent GPP at our sites and this was also reflected in the biomass exports.While the number of "grazing" events were similar at both Sites A d and A s (5 in Year 1 and 4 in Year 2), the higher productivity in the deep-drained site was reflected in larger C biomass exports (30 % more in Year 1 and 35 % more in Year 2 compared to site A s ).N biomass exports were also on average 36 % higher in Site A d over the two years.By comparison, eight "grazing" events were possible in the nutrientrich Site B. Management practice in the form of grazing regime would therefore have a greater impact on GPP than at the other sites.

Respiration drivers
Modelled annual R eco was significantly different among all sites and was highest at the nutrient-rich site (B) with an estimated mean value of 2322 g C m −2 yr −1 , which is similar to values measured for grasslands over organic soils in Europe (Gilmanov et al., 2007;Elsgaard et al., 2012;Beetz et al., 2013).R eco significantly decreased with increasing soil moisture and decreasing soil temperatures in accordance with other studies on managed organic soils (Teh et al., 2011;Elsgaard et al., 2012).Soil temperatures were not significantly different among all sites and therefore did not explain the differences in R eco .VMC varied widely at the nutrientrich site and wet-dry cycles can enhance peat degradation and therefore increase soil respiration (Aerts and Ludwig, 1997;Chow et al., 2006) as was also reflected in the higher DOC fluxes measured at that site.Furthermore, as high GPP would also result in high autotrophic respiration rates, the sustained high respiration rates at Site B, especially at the end of the growing season, might be the result of enhanced levels of heterotrophic soil respiration in the rhizosphere, known as the "priming effect" (Kuzyakov, 2006) driven by the addition of labile C in the form of root exudates and litter.
The two well-drained organic soils (nutrient-poor Site A d and nutrient-rich Site B) contrasted in that, while Site A d was less productive (20 % lower) due mainly to a lower nutrient content, it also displayed reduced R eco sustained through the year (30 % lower on average).Lower rates of autotrophic respiration due to lower productivity could only explain the difference during the growing season.Given that they display similar soil temperature profiles, other edaphic characteristics may be at play.When peat soils are drained, organic matter can be rapidly oxidised by aerobic metabolic processes.However, rates of aerobic decay vary widely across drained organic soil types even under conditions of similar climate, water table and management (Byrne et al., 2004;Couwenberg, 2011).The lower organic matter (OM) content (30 and 70 % at Site A d and B respectively) may have influenced respiration rates although Leiber-Sauheitl et al. (2013) found that CO 2 emissions were not affected by lower soil organic carbon content in the surface horizon of cultivated histic soils.Leifeld et al. (2012) demonstrated that the quality of the OM at the surface of organic soils that have been drained for a long time is poor (e.g.lower polysaccharides content) and is less sensitive to C loss via heterotrophic respiration due to preferential metabolism of labile organic matter fractions.While OM quality was not specifically examined in this study, the soil C : N ratios of the deep-drained sites were very similar (Table 1) suggesting similar OM quality.However, the depth of OM already exposed to oxygen would have been greater over time at Site A d as this site has been drained for > 100 years so that the recalcitrant OM fraction may be greater compared to Site B. Heterotrophic metabolism can also be limited by nutrient availability, so that despite the similar site characteristics and soil C : N, the greater soil nutrients and higher pH at Site B may explain the greater respiration at this site (Table 1).Indeed the elevated respiration rates are consistent with the greater soil nitrate concentrations at this site, which may partly reflect elevated nitrification rates but also that this site receives low inorganic fertiliser inputs.It should be noted that while marl is present under the peat at Site B, the contribution of the abiotic CaCO 3 -derived CO 2 effluxes to the annual soil CO 2 fluxes is deemed marginal (Kuzyakov, 2006).
In this study, respiration rates contrasted between the two nutrient-poor sites where higher water table levels led to decreased respiration rates, which have been observed elsewhere (Leiber-Sauheitl et al., 2013;Skiba et al., 2013).The linear response of R eco to water table depth as modelled by Eq. ( 2) and the good fit between modelled and observed data for the derived models (Fig. 4) indicated that R eco was not restricted by very dry hydrological conditions.This is due to the relatively constant high moisture content (Irish soils are rarely subject to moisture deficit) linked to a high precipitation regime with the number of wet days well distributed throughout the year (Xu et al., 2004;Lawton et al., 2006).Soil temperature also exerted a strong influence on R eco , as has been found in other studies and this explained the interannual variation of annual R eco at this site.Respiration rates were reduced by 12 % at both sites A s and A d in Year 2 which had similar water table levels but colder soil temperatures (at 5 cm, Fig. 2) than Year 1.Interestingly, productivity was less affected by the cooler temperatures, suggesting that soil metabolism was more affected by this climatic feature than photosynthetic activity.Similarly, higher water table levels affected R eco more strongly than GPP (through LAI) with 27 % lower annual R eco in the shallow-drained site in both years compared to the deep-drained site.Taken together our results demonstrate that low respiration rates controlled by higher water table levels and colder soil temperatures were the main drivers of NEE at the nutrient-poor sites.Thus, while the CO 2 -C sink potential of these sites may be threatened by climate change, it may be safeguarded by ensuring a shallow drainage status (water table remaining above −25 cm).

Magnitude of CH 4 and N 2 O fluxes
Uncertainty associated with both CH 4 and N 2 O fluxes stems from source variation (temporal and spatial) as well as limitations in measurement and methodology.As CH 4 production is expected to occur mainly within the anoxic soil horizons, i.e. below the groundwater table, the absence of CH 4 emissions at Sites B and A d are consistent firstly with annual mean water table depths (below −30 cm) and secondly with the absence of aerenchymous plant species that can transport CH 4 from the soil to the atmosphere (Couwenberg, 2009).Fluxes close to zero were also recorded from drained grasslands over peat in Germany (Drösler et al., 2013;Leiber-Sauheitl et al., 2013) andDenmark (Petersen et al., 2012).
The estimated mean annual CH 4 fluxes for the shallowdrained nutrient-poor Site A s (18 ± 15 kg CH 4 ha −1 yr −1 over 2 years) are at the lower range of values reported for wet grasslands in Germany (Leiber-Sauheitl et al., 2013) and are only 10 % of values reported for managed fen meadow and extensively managed grassland on peat in the Netherlands (Kroon et al., 2010;Schrier-Uijl et al., 2010).Our sampling frequency may have under-estimated annual emissions as some fluxes might have been missed during wetter periods.However, measurements for CH 4 were carried out at all sites over a wide range of water table levels or VMC.The absence of CH 4 fluxes recorded during the wetter periods in Site B in particular may be due to several reasons.Since net CH 4 fluxes are measured, CH 4 oxidation and consumption may be taking place within the uppermost well-aerated horizon.Suppressed methanogenesis may also occur in the deeper layers due to the abundance of electron acceptors other than oxygen, such as nitrate (Table 2).The low level or even absence of CH 4 fluxes may also be due to the relatively shallow peat depth (< 1 m) as suggested by Levy et al. (2012) and the fact that these sites have been drained for a very long time (Flasse et al., 1998).Poor quality and supply of substrates may also restrict methanogenesis (Couwenberg, 2009).Given that a recent meta-analysis has demonstrated that vegetation is the strongest explanatory parameter for CH 4 flux variability (Levy et al., 2012), it is likely that the higher CH 4 fluxes measured from the shallow-drained nutrient-poor sites were also a product of the presence of Juncus effusus (aerenchymous species) as seen in other studies (Petersen et al., 2012;Herbst et al., 2013).Overall, our results demonstrate that CH 4 fluxes were negligible, if not absent components of the C balance of extensively drained grasslands over organic soils in a maritime temperate climate, and that CH 4 was emitted only in very small amounts when the mean annual water table was around −23 cm.
The absence of N 2 O emissions from the nutrient-poor sites was consistent with the lack of artificial fertiliser use for several decades at these sites, and similar results have been reported from an extensive drained moorland in Scotland (Skiba et al., 2013).In nutrient-rich sites, the potential for N 2 O emissions is much higher, and is additionally forced by fertiliser applications to these more productive systems.Estimated annual N 2 O emissions from Site B are nonetheless still low by comparison with most grasslands situated over peat (closer to the IPCC Tier 1 default values for shallow-drained nutrient organic soils).This study provides further evidence that N 2 O fluxes are predominantly affected by the grassland management type, which is not represented in the categories with the IPCC guidelines (2014).Most investigations are carried out in intensively managed grasslands.While being nutrient-rich, Site B is still very much "extensive" given the low artificial N input.Our values are, however, comparable to Dutch figures reported by Langeveld et al. (1997) andvan Beek et al. (2011) where grasslands had typically less fertiliser and manure inputs over time.Overall, our results are not sufficient to conclude firmly on the insignificance of N 2 O fluxes in extensive grassland over organic soils and further investigations in this area are warranted to ascertain the potential for large denitrification losses.Since N 2 O uptakes were detected on occasion (Fig. 9), N 2 O reductase activity could be high and therefore large dinitrogen (N 2 ) emissions may be possible.This study also highlights the need for further research on intensive measurement of N 2 O based on C : N ratios, which may not reflect soil organic C degradation as seen in the contrasting ecosystem respiration measurements from our two sites with similar C : N ratios.

Fluvial C losses
Studies have shown that fluvial C losses can be large components of terrestrial C balances at field and landscape scales, particularly where organic soils predominate (Worrall et al., 2009;Dinsmore et al., 2010;Buffam et al., 2011).Furthermore, the importance of this component can be magnified by management practices given that large increases in DOC export (51 to 118 %) can occur following drainage of such soils, related to greater DOC production under aerobic conditions and soil redox conditions during wet and dry periods (Wallage et al., 2006;Clark et al., 2009;Urbanová et al., 2011).In this study, DOC fluxes were greatest for the drained nutrientrich site but lower than values reported for drained peatland catchments in the same climatic region (Gibson et al., 2009;Rowson et al., 2010).More significantly, the 2-year mean DOC flux from the nutrient-poor site (11.5 g C m −2 yr −1 ) is at the low end of values reported for temperate semi-natural peatland catchments in Ireland and the UK (Dawson et al., 2004;Dinsmore et al., 2010;Koehler et al., 2011;Kiely et al., 2014).This suggests that DOC fluxes from nutrient-poor organic soils under extensive grasslands in Ireland may be more similar to semi-natural ecosystems.Annual DOC fluxes from the nutrient-poor site were almost 4 times lower in Year 2 reaching a low value of 4.9 g C m −2 yr −1 .Inter-annual variability in DOC flux at this site is likely due to reduced precipitation and greater evapo-transpiration during the second hydrological year (runoff 897 and 745 mm in Years 1 and 2 respectively).However, the decline in DOC export from Year 1 to 2 was partially offset by a greater flux of aqueous CO 2 export, implying that drier conditions were more conducive to soil OM and soil pore water OM mineralisation.In both years, the flux of DIC (derived from soil respiration) exceeded the flux of DOC, and was predominantly in the form of CO 2 .The high pCO 2 of samples implies that accurately accounting for aqueous CO 2 fluxes requires that samples are taken close to source to account for C that is otherwise rapidly evaded to the atmosphere, particularly with waters with low alkalinity.While the literature on aqueous inorganic C fluxes from agricultural catchments is not extensive, our observations are consistent with findings for small boreal streams where exports of CO 2 were also substantial and atmospheric evasion of CO 2 from streams was equivalent to approximately half of the total stream C flux (Wallin et al., 2013).Our findings also provide support for arguments for greater incorporation of stream C fluxes to improve regional and global C accounting (e.g.Downing et al., 2012).At Site B the DIC flux was also considerable, but comprised a lower fraction of the total.Notably this DIC export was predominately in the form of bicarbonate so that the soil respiratory origins of much of this fraction are indicated to behave more conservatively during downstream transport but nevertheless remains an important term in the terrestrial C balance.These results support an argument for the requirement to report other fluvial C components in GHG inventories, notably DIC and POC (e.g.Hope et al., 2001;Dinsmore et al., 2010).

Complete NECB
Concurrent GHG and fluvial C balance studies on organic soils have been undertaken on only a few sites worldwide and this is the first full NECB of a temperate drained organic soil under grassland.At the nutrient-rich site, NECB was relatively high (663 ± 42 g C m −2 yr −1 ) considering that the grassland is extensive and this value is greater than those reported for intensive grassland on peat in both Germany (Beetz et al., 2013) and the Netherlands (Veenendaal et al., 2007).However, the value falls within the range of annual emission rates of 410-760 g C m −2 yr −1 reported by Couwenberg (2011), who reviewed several temperate drained peat soils under grassland.It should be noted, however, that the values above do not include fluvial C losses, which in this study accounted for 11 % of the total NECB for the nutrient-rich site.While annual biomass exports were at the low end of literature values, e.g.317-515 g C m −2 yr −1 (Veenendaal et al., 2007;Beetz et al., 2013), they accounted for 53 % of the NECB (Table 4).The 2-year mean NECB of the deep-and shallow-drained nutrient-poor sites were considerably smaller than the nutrient-rich site (40 and 16 % respectively).Biomass exports were also the biggest components of the NECB at these sites (Table 4) and were sufficient to influence the source-sink balance of the shallow-drained site in both years and at the deep-drained site in Year 2 (i.e.shift from a NEE sink to a net source of C).Given the impact of the biomass removal on NECB, controls on grazing regimes could therefore significantly influence the C sink capacity of these ecosystems, as documented for an abandoned former pasture over peat in the Netherlands (Hendricks et al., 2007).The NECB of the shallow-drained nutrient-poor site is indeed consistent with values reported for an extensive grassland in Germany, which was rewetted 4 years previously (Beetz et al., 2013).However, in the case of nutrient-poor organic soils, fluvial C losses were not a negligible component of the overall NECB and could easily influence the final C balance of these ecosystems.

Implications for reporting and climate change mitigation strategies
NEE added to the C losses from the removal of biomass (assuming instantaneous emission of cut biomass) and averaged over a multi-annual measuring period, gives a best-possible estimate of emissions/removals and therefore the most accurate CO 2 -C emission factor (EF) (Couwenberg, 2011;Elsgaard et al., 2012).The EF for the nutrient-rich deep-drained grassland (Site B) was 5.84 t CO 2 -C ha −1 yr −1 which is in the range of values given in the IPCC guidance (2013): 6.1 (5.0-7.3)t CO 2 -C ha −1 yr −1 for this land use category.It corresponds well with Tier 2 country-specific EFs for permanent grassland over organic soils: 5 t CO 2 -C ha −1 yr −1 in Germany (Federal Environmental Agency, 2013), 5.19 t CO 2 -C ha −1 yr −1 in the Netherlands (Netherlands Environmental Assessment Agency, 2009) and 5.17 t CO 2 -C ha −1 yr −1 in Denmark (National Environmental Research Institute, 2013).
In contrast, our nutrient-poor sites produced much lower CO 2 EFs of 2.35 t CO 2 -C ha −1 yr −1 in the deep-drained and 0.62 t CO 2 -C ha −1 yr −1 in the shallow-drained.These figures are outside the lower range of EFs for temperate drained nutrient-poor sites in the IPCC guidance (2014): 5.3 (3.7-6.9)t CO 2 -C ha −1 yr −1 .This default value is not further stratified between shallow and deep-drained grasslands and is based on only seven identified study sites, six of which are from Germany (Drösler et al., 2013).Nonetheless, the multisite GHG investigation reported by Drösler et al. (2013) includes nutrient-poor sites (extensive grasslands), which were a CO 2 -C sink (NEE minus biomass removal) in some years.In Denmark, an EF of 1.25 t CO 2 -C ha −1 yr −1 is applied to non-fertilised permanent grassland.Along with the results presented here, this supports our premise highlighting considerable heterogeneity in the C balance of grasslands on or-ganic soils within the European temperate zone.While the stratification of default CO 2 EFs according to nutrient and drainage status is indeed supported by this study, variability in the impact of these factors also reflects the historical and contemporary management practices therefore advocating a progression towards the Tier 2 reporting level for countries with significant areas of organic soils under grassland.
The EF for CH 4 at both the deep-drained sites (regardless of nutrient status) was zero and for the shallowdrained nutrient-poor grassland was calculated at 18 kg CH 4 ha −1 yr −1 .The absence of CH 4 fluxes from both our well-drained organic soils is at variance with the IPCC guidance which presents an EF of 16 kg CH 4 ha −1 yr −1 for nutrient-rich and 1.8 kg CH 4 ha −1 yr −1 for nutrient-poor.This over-estimation could be compensated by using the EF for nutrient-poor drained organic soils in the case of shallowdrained nutrient-poor sites.If water table status is used as a primary factor, the only default EF provided is for shallowdrained nutrient-rich at 39 kg CH 4 ha −1 yr −1 which would also lead to a large over-estimation of CH 4 fluxes.
The EF for N 2 O of zero at the nutrient-poor site and 1.6 kg N ha −1 yr −1 at the nutrient-rich site are both very low compared to the default EF in the IPCC guidance (2014).However, our value is likely to be conservative as the fertilisation events were not included in this figure.Despite being subject to high uncertainty, our results suggest again the likely significance of management system, especially with regard to fertilisation and stocking regime, factors which are currently missing from Tier 1 methodology.
In order to calculate an EF for DOC, it is critical to account only for the portion of DOC that could be completely mineralised and re-emitted to the atmosphere.While lake sediments can be sites for significant permanent burial of terrestrially fixed C, in most river-lake catchments mineralisation far exceeds burial (Algesten et al., 2003), and indeed studies have shown that most of the DOC flux from drained peat catchments can be mineralised (Kölher et al., 2002;Jones et al., 2013).Despite such observations, the oceans still receive considerable inputs of terrestrial organic matter (ca.0.4 Gt yr −1 according to Richey, 2004) although the very minor amounts of unambiguously terrestrial material identified in seawater and marine sediments imply that the vast majority is re-mineralised (Bauer and Bianchi, 2011).As per the updated IPCC guidance, a 90 % fraction was used to calculate an overall EF for DOC of 22.1 t C ha −1 yr −1 , being much lower at the nutrient-poor site (10.3 t C ha −1 yr −1 ) compared to the nutrient-rich site (33.9 t C ha −1 yr −1) .The latter value is within the range of the default EF for temperate drained organic soils: 31 (19-46) g C m −2 yr −1 (IPCC, 2014).Given the high inter-annual variability experienced at the nutrientpoor site, further long-term study would be warranted in order to support the development of a country-specific EF which would include a "nutrient status" stratification.
Overall, the current default EF may not be representative of the variety of grasslands over organic soils present in the www.biogeosciences.net/11/4361/2014/Biogeosciences, 11, 4361-4379, 2014 maritime fringes of Western Europe.As only 1 full year was monitored in the nutrient-rich Site B and 2 years at Site A, a longer monitoring period to increase the range of climatic conditions and additional direct measurements from other regionally representative drained organic soils would help attain more robust emission factors by averaging inter-annual variability.In turn, this would help to implement more effectively Tier 2 methodologies for IPCC inventories.In the context of the new opportunity given by the IPCC 2013 Wetland Supplement to report rewetted drained organic soils, this study underpins the need to investigate the rewetting of high C content nutrient-rich organic soils under grassland in Ireland as a priority.The sustained high decomposition rates recorded at the nutrient-rich site could only be compensated by reducing microbial oxidation of peat via higher water table levels.Rewetting of these site types should therefore be a priority of any climate change mitigation strategy focusing on decreasing C losses from soils in Ireland.Furthermore, the neutral or C sink capacity of nutrient-poor organic soils may also be strongly affected by climate change.With higher predicted temperatures affecting peat decomposition, respiration rates might increase faster than photosynthetic assimilation due to nutrient limitation.Predicted higher precipitation for the studied region could however work in synergy with government-led actions to (1) introduce agricultural schemes to maintain a relatively high water table (above −25 cm); and (2) reduce the grazing regime to actively promote C uptake by these ecosystems.

Concluding remarks
This study has provided estimations of GHG fluxes and waterborne C losses from contrasting grasslands over organic soils and has shown that these reflect not only regional variability but also local management practices.Hence, despite the critical role of nutrient and drainage status, detailed reporting of grassland management type may provide significant additional information for more accurate regional extrapolation of C fluxes.The large difference in net C balances between the main locations stems mostly from (1) NEE estimations and (2) biomass exports, which are affected by the grazing regime.NEE estimations were driven mainly by local climate, soil fertility, water table level and potentially soil organic matter quality.These attributes are in turn intimately linked to past and current management practices in terms of drainage duration and intensity and inputs.Most Irish grasslands over organic soils were drained several decades ago and are managed as extensively grazed pastures with little or no fertilisation.Nutrient-poor sites in Ireland may not exert appreciable impacts on the atmosphere in terms of NECB, particularly where the mean annual water table remains within −25 cm of the soil surface.In contrast, grassland over nutrient-rich, drained organic soils (with higher C content) produce much higher GHG emissions and also rep-resent hotspots for fluvial C losses.As the C balance of such sites is driven to a large extent by soil respiration, the inverse relationship of this parameter with soil moisture suggests these sites should be targeted for rewetting as a strategy to mitigate C emissions.This comprehensive investigation supplied effective information for Ireland to refine emission factors currently proposed for Tier 1 drained organic soils.Longer monitoring is required to capture inter-annual variability which affects NEE in particular (being the small difference between two large terms GPP and R eco ) (Drösler et al., 2008).While a "grassland type" stratification, representing various management systems, may not be warranted at international guidelines levels, national research should be extended to other sites to provide more accurate emissions from extensive drained organic soils pastures and help Ireland progress towards Tier 2 methodologies.

Figure 2 .
Figure 2. (a) Mean monthly Photosynthetic Photon Flux Density (PPFD) (µmol m −2 s −1 ) at Site A (April 2011 to March 2013) and Site B (April 2011 to March 2012) and (b) mean monthly soil temperature at 5 cm depth (T 5 cm ± standard deviation) at Site A d , Site A s and Site B.
Figure 3. (a) Mean monthly water table levels (cm) at Site A d and A s (error bars are standard deviation of the monthly means) and (b) Volumetric Moisture Content (% VMC) at Site B (in the surface 5 cm of the peat).

38Fig. 4 .
Fig. 4. Relationship between observed and modelled Gross Primary Production (GPP) and Ecosystem Respiration (Reco) fluxes (mg CO2 m -2 h -1 ) for Sites A and B. Dark circles indicate data used in the construction of models and open circles indicate independent test data.

Figure 4 .
Figure 4. Relationship between observed and modelled Gross Primary Production (GPP) and Ecosystem Respiration (R eco ) fluxes (mg CO 2 m −2 h −1 ) for Sites A and B. Dark circles indicate data used in the construction of models and open circles indicate independent test data.

Figure 5 .
Figure 5. Monthly mean gross primary production (GPP), ecosystem respiration (R eco ) and net ecosystem exchange (NEE) (g CO 2 -C m −2 month −1 ).Error bars represent ± standard deviation.n = 7 for Site A d , n = 5 for Site A s and n = 9 for Site B. Negative NEE values indicate that the site was a net sink for CO 2 -C for that month.

Figure 6 .
Figure 6.Relationship between mean annual water table (cm) and mean annual Leaf Area Index (LAI) at the nutrient-poor sites A d and A s (Year 1 and 2 combined).

Figure 7 .
Figure 7. Vegetation mean growth rate (cm day −1 ) at Sites A d and A s during periods between simulated grazing events (cuttings) during Year 1 and 2 (vertical bars denote ± one standard error; n = 7 for Site A d and n = 5 for Site A s ).Means with the same letter are not significantly different (p < 0.001).
Ireland with warm winters (very few ground frosts) but cool summers.Site B in Lanesborough, Co. Longford (latitude: 53 • 39 N, longitude: 7 • 56 W) is also low-lying, situated at 38 m elevation but is typical of the midlands landscape, with late spring frosts occurring as late as the beginning of June.Despite comparable long-term annual mean temperatures, lower minimum and higher maximum temperatures are typically recorded at Site B. A minor gradient in precipitation is also present with higher mean rainfall in Site A (Table1).Potential evapo-transpiration in that region is estimated at 32-36 % of annual precipitation (estimated long-term annual runoff of ca.700 mm), compared to 44 % (annual runoff, ca.586 mm) for Site B (

Table 2 .
Soil and land use characteristics of the research sites.

Table 3 .
Estimated parameter values for gross photosynthesis (GPP) and ecosystem respiration (R eco ) models (see related equations in the text).Standard error of the model parameter in parentheses.Coefficient of determination (r 2 ) values and equation number are shown.p-values for all parameters < 0.005.

Table 4 .
Annual fluxes of ecosystem respiration (R eco ), gross primary production (GPP), net ecosystem exchange (NEE) of CO 2 , CH 4 , dissolved organic carbon (DOC), particulate organic carbon (POC), excess CO 2 and bicarbonate (HCO −3 ) together with biomass exports, CH 4 emissions from enteric fermentation and net ecosystem carbon balance (NECB).Standard errors are shown in parentheses for all gas fluxes R eco , GPP, NEE and for fluvial fluxes.One standard deviation is shown in parentheses for all other components.Site A d = nutrientpoor deep-drained, Site A s = nutrient-poor shallow-drained and Site B = nutrient-rich drained.Positive values indicate a loss of carbon from the site and negative values indicate an uptake of carbon by the site.