An integrated assessment of nitrogen source, transformation and fate within an intensive dairy system to inform management change

From an environmental perspective optimised dairy systems, which follow current regulations, still have low nitrogen (N) use efficiency, high N surplus (kg N ha-1) and enable ad-hoc delivery of direct and indirect reactive N losses to water and the atmosphere. The objective of the present study was to divide an intensive dairy farm into N attenuation capacity areas based on this ad-hoc delivery. Historical and current spatial and temporal multi-level datasets (stable isotope and dissolved gas) were combined and interpreted. Results showed that the farm had four distinct attenuation areas: high N attenuation: characterised by ammonium-N (NH4+-N) below 0.23 mg NH4+-N l-1 and nitrate (NO3--N) below 5.65 mg NO3--N l-1 in surface, drainage and groundwater, located on imperfectly to moderately-well drained soils with high denitrification potential and low nitrous oxide (N2O) emissions (av. 0.0032 mg N2O-N l-1); moderate N attenuation: characterised by low NO3--N concentration in drainage water but high N2O production (0.0317 mg N2O-N l-1) and denitrification potential lower than group 1 (av. δ15N-NO3-: 16.4‰, av. δ18O-NO3-: 9.2‰), on well to moderately drained soils; low N attenuation—area 1: characterised by high NO3--N (av. 6.90 mg NO3--N l-1) in drainage water from well to moderately-well drained soils, with low denitrification potential (av. δ15N-NO3-: 9.5‰, av. δ18O-NO3-: 5.9‰) and high N2O emissions (0.0319 mg N2O l-1); and low N attenuation—area 2: characterised by high NH4+-N (av. 3.93 mg NH4+-N l-1 and high N2O emissions (av. 0.0521 mg N2O l-1) from well to imperfectly drained soil. N loads on site should be moved away from low attenuation areas and emissions to air and water should be assessed.


Abstract
From an environmental perspective optimised dairy systems, which follow current regulations, still have low nitrogen (N) use efficiency, high N surplus (kg N ha -1 ) and enable ad-hoc delivery of direct and indirect reactive N losses to water and the atmosphere. The objective of the present study was to divide an intensive dairy farm into N attenuation capacity areas based on this ad-hoc delivery. Historical and current spatial and temporal multi-level datasets (stable isotope and dissolved gas) were combined and interpreted. Results showed that the farm had four distinct attenuation areas: high N attenuation: characterised by ammonium-N (NH 4 + -N) below 0.23 mg NH 4 + -N l -1 and nitrate (NO 3 --N) below 5.65 mg NO 3 --N l -1 in surface, drainage and groundwater, located on imperfectly to moderately-well drained soils with high denitrification potential and low nitrous oxide (N 2 O) emissions (av. 0.0032 mg N 2 O-N l -1 ); moderate N attenuation: characterised by low NO 3 --N concentration in drainage water but high N 2 O production (0.0317 mg N 2 O-N l -1 ) and denitrification potential lower than group 1 (av.

Introduction
Reactive nitrogen (N) surplus in agricultural watersheds accumulates in the rooting zone of soils due to low nutrient use efficiencies [1]. This N storage can affect connected water quality for long periods where the biogeochemical time lag is long [2]. Therefore, minimising the N load in areas with high biogeochemical time lags and connectivity to water bodies is an important measure for N attenuation. Nitrogen loading originates from a mixture of animal waste and fertiliser inputs, comprising both inorganic (urea, calcium ammonium nitrate (CAN)) and/or organic (dairy soiled water (DSW), manure, slurry and urine) components. As the subsurface is heterogeneous surplus N can be transformed at different rates through mainly biological processes within the N cycle and especially nitrification and denitrification, in which ammonium (NH 4 + -N) is oxidised to nitrate (NO 3 --N) and then reduced to di-nitrogen gas (N 2 ) [3]. This and other pathways (e.g. nitrification, DNRA (dissimilatory nitrate reduction to ammonium), anammox) are composed of sequential reactions, with the production and possible release of intermediate and undesirable N compounds to the environment, e.g. NO 3 --N, nitrite (NO 2 --N) and nitrous oxide (N 2 O). Current policy instruments, under the EU Water Framework Directive [4] treat the farm as a homogeneous soil block where there is no acknowledgement of the ad-hoc delivery of direct and indirect reactive N losses to water and the atmosphere. However, the ability of the soil/subsoil and underlying geology to attenuate this N surplus is heterogeneous. Dividing an intensive dairy farm into N attenuation capacity areas based on actual ad-hoc delivery could minimise future losses of N. Clagnan et al. [5,6] studied N attenuation from plot to field scales using biogeochemical, gas and isotope techniques, ranking poorly drained sites (artificially drained and un-drained) with respect to their natural attenuation capacity. Results showed that deep groundwater drainage systems did not negatively affect the N attenuation capacity of a soil. A concept of "net" water origin, N source, transformation and fate was developed using water samples from borehole screen intervals and end-of-pipe locations [5,6]. This concept enables interpretation of water origin and identifies a clear migration pathway for that signature (e.g. precipitation, groundwater and a mixed enriched signal inferring migration along a pipe or ditch) over much larger areas, and thus highlights connectivity or dis-connectivity of N sources, transformations, pathways and receptors. Until now this concept has never been examined at farmscale (125 ha).
Research on soil functions has indicated that the N water purification function of soil, identified by Schulte et al. [7] as "the amount of denitrification required to ensure that the N surplus leaving the rooting zone does not lead to excess groundwater N concentrations", is higher in poorly drained (low permeability) and lowest in freely drained (high permeability) soils [8,9]. The N water purification function of different agricultural soils has a significant impact on reducing/controlling N concentration and the ratio among N species, with denitrification being the main process of N removal [10,11]. N speciation and attenuation capacity is regulated by many other environmental factors, e.g. substrate concentrations, plant coverage, management and weather [12].
The role of artificial drainage on N transfer, transformation and migration on intensive dairy farms with variable soil types and drainage classes remains unclear. A drainage pipe installed within the sub-soil is likely to connect areas of "low" and "high" denitrification potential, depending on soil functional characteristics [7]. The end-of-pipe water sampled reflects the composite attenuation capacity of the subsoil draining into this pipe. However, these systems may reduce N transformation potential by creating unsuitable conditions for in situ denitrification (e.g. higher oxygen concentration, lower water saturation and shorter residence time), leading to greater NO 3 --N losses. Equally, zones of high soil N attenuation capacity may be by-passed by the drainage system. Other N species, such as NH 4 + -N, NO 2 --N and N 2 O, are seldom considered and there is a need to examine if drainage system pathways in heterogeneous soils can utilise areas which support N attenuation capacity, or by-pass and nullify this capacity, with increased reactive N losses from the system. There have been limited investigations of NO 3 --N distributions in shallow groundwater systems under such intensive dairy farms [8,13]. On the present study site, Baily et al. [14] used NO 3 --N natural isotopic abundances for three sampling periods (April, August and Dec) to deduce the role of denitrification in soil N dynamics. Other studies [10,15] estimated such N-losses to be approximately 106 kg N ha -1 , explained by hydrological and geochemical factors (e.g. availability of dissolved oxygen (DO) concentration and redox potential (Eh)). While these studies offered insight into the fate of NO 3 --N in shallow groundwater, they did not increase knowledge pertaining to N provenance, multi-level spatial distribution or the transformation of N along shallow pathways and deep pathways or the ultimate fate of the lost N. Nutrient concentrations, hydrochemical parameters (e.g. pH, DO, Eh) and soil properties (e.g. saturated hydraulic conductivity, k s ) can provide qualitative evidence of spatial and temporal variation in N-transformation processes and the soil conditions supporting these. Additional gas analyses may increase the understanding of local N-transformation processes, which can be quantified (in term of sinks and sources of N 2 O) by the release of their final gaseous products (i.e. N 2 and N 2 O). However, since multiple N-transformation processes can contribute to N 2 and N 2 O production, further analysis using isotopic abundances can help elucidate the different N sources and transformation processes (e.g. [16]).
The objective of the present study was to examine how dairy farm N surplus, N source and N attenuation on a heterogeneous landscape can be managed to reduce N loads to vulnerable areas, thereby protecting water quality into the future. To meet this objective, an extensive dataset was assembled over a 12-year period, from a combination of previous and new fieldwork.

Study site
The Teagasc Johnstown Castle, Co. Wexford, (52˚17'30@N, 6˚29'50@W) (Fig 1) study site comprises two units of grassland, one of 72.9 ha, termed "down-gradient", and an "up-gradient" of 50.8 ha separated by a roadway. The up-gradient subsurface drainage system is connected to the down-gradient system through an underground connector pipe. The farm, considering the two grassland units together, is intensive and operates at 3.1 Livestock units per hectare (LU ha -1 ). Nitrogen inputs arise from urea (spread February to April), calcium ammonium nitrate (CAN) (May to September) and manure (spring) for a total of 257 kg N ha -1 inputs of inorganic fertiliser and 103 kg N ha -1 of organic (5 years average). The central area of the dairy farm receives dairy soiled water (DSW, consisting of rainwater, yard and milk parlour washings) from Feb to Oct through a rotational irrigation system (Fig 1, former spreading area: plot with locations 11 and 14; current spreading area: between the Met station and location 19). To quantify the N surplus that could affect the drainage system, farm partial N balances for each year were calculated as per Treacy et al. [17], utilising stocking rates, N inputs (inorganic and organic fertilisers and concentrate feed (volume and composition)) and N exports (milk production (volume and composition) and slurry). Cows were milked twice daily (07�30 h and 15�30 h), with milk yield (kg) registered for each cow each time. The milk composition (fat, protein and lactose concentrations) for each cow was tested on a successive morning and evening every two weeks using a Milkoscan 203 (Foss Electric DK-3400, Hillerød, Denmark). Milk solids were calculated using the method of Tyrell and Reid [18]. For both milk and concentrates fed the N value is averaged across the farm. This region has an average annual air temperature of 10˚C (1981-2011), average annual rainfall (1981-2011) is 1037.5 mm with maximum intensity between Sep and Nov. There is an in situ synoptic meteorological station (Fig 1) on site which records daily rainfall, wind speed and hours of sunshine. This data was inputted into the grassland hybrid soil moisture deficit (HSMD) model of Schulte et al. [19] to estimate effective drainage (ED, mm day -1 ). To examine differences in ED modelling was conducted for well-, moderately-and poorly-drained soils using data from 2008 to 2017. Soil texture varies from fine loam to clay loam (Brown Earth, Gleyic Cambisol with Irish Sea till origin [20], Fig 1), with small areas of sandy textured soils. Subsoil is of moderate permeability (0.2-10 m; 5 x 10 −8 m s -1 <k s < 5 x 10 −4 m s -1 ) but can be interspersed with high permeability gravel and/or sand lenses [8,21]. Unsaturated zone hydrologic time lag is from 1 (shallowest areas) to 3 (deepest areas) years with probability analysis indicating 1.5 years 85% of the time [14], presenting a mismatch between best management practice at farm level, when nutrients are lost, stored and mineralised and when leaching affects N concentrations along subsurface pathways. The average water table depth on site from 2005-2014 is 2.8 (±1.7) m below ground level (bgl) (deepest from July to Sep i.e. 3.0 m bgl and shallowest from December to February i.e. 2.3 m bgl). The low permeability bedrock is Pre-Cambrian greywacke mixed with quartzite at a depth of 10-12 m (k s , 3.6 x 10 −6 m s -1 ), containing a poorly productive aquifer, which is classified as receiving 1 to 50 mm yr -1 [14].

Surface and subsurface monitoring network
A large surface and subsurface monitoring network was collated in GIS and a map developed for the purpose of the study. The open ditch network and groundwater monitoring components of the system were well documented, but the subsurface drainage system remained unmapped. Examination of historical maps and discussions with farm staff identified likely positions of in-field drains, which were confirmed by field work (Fig 1, S1 Table). Both upand down-gradient units contain poorly-imperfectly drained soils (79% and 28%, respectively) (Fig 1). These areas are all artificially drained by an in-field pipe network of 10.1 km (Fig 1) composed of either corrugated slotted plastic (80 or 100 mm at 0.5-1 m depth) or concrete non-perforated pipes (600 mm at 1 m depth act as connectors).
The components of the surface and subsurface drainage system are as follows: 1) Up-gradient component: a herring bone drainage network (80 mm, variable depth) conveys drainage water from an area of 24 ha, to an underground outlet (D2, Fig 1) and joins with another outlet at D3, which passes to a junction also fed by a drainage system passing through D1. This composite migrates in a fully cased concrete pipe and eventually discharges into the open ditch in the down-gradient unit. 2) Down-gradient component including open ditch: the subsurface drain that transfers water from locations 2, 3, 7, 8, 9 and 10 is an ad-hoc drain developed over time with many unknown extensions and discharges to the start of the open ditch. A short herring bone system discharges water into the open ditch adjacent to location 14. The open ditch extends from 11 to D4, before being piped underground to the outlet at D8 (Fig 1). There is an access point in the underground section at D5-6-7. Individual drains connect to the underground part of this primary drain at location 30 (fully cased draining a marl pond) and 36. 3) Unconnected components: an in-field herring bone drain in the area of 15, 16 and 17 flows in the opposite direction to the open drain and has an offsite outlet, as does the system around 19 and 20. Any discharge from the nearby artificial lake system (D9) does not affect D8. Prior to the 2014 sampling campaign, water sampling locations were added to this system, i.e. three positions along the subsurface drainage system (D1, D2, D6), one position along the surface stream (D5) and three within the groundwater network (1, 22, 23) (Fig 1).

Historical data
Twelve years of data for the down gradient dairy farm were collated from the groundwater monitoring network (S1 Table, No 5). The groundwater monitoring system consists of three components: 1) five sets of multilevel boreholes (Fig 1) representing three depths: subsoil (4.0-7.5 m bgl), bedrock (16.8-23.0 m bgl) and the subsoil-bedrock interface (11.0-13.0 m bgl); 2) a network of piezometers and boreholes (Fig 1), with screen depths from 1.95 to 8.95 m bgl installed to sample shallow groundwater within the subsoil layer; 3) a single borehole drilled to 37 m bgl was installed to sample deep groundwater (18, Fig 1).
Fieldwork was conducted from Dec 2005 to June 2017 (S1 Table,

Current data
Fieldwork was undertaken in Sep 2014 and June 2017 to collect water samples for all locations in Fig 1. These two sampling campaigns were carried out to integrate different techniques (i.e. stable isotope and dissolved gas analyses (only for Sep 2014) and all elements of the water continuum (i.e. shallow and groundwater, in field drainage and open ditches). These campaigns also deduced the stability of the N signatures and dissolved gases on site within shallow water which was highlighted by Baily et al. [14] and Jahangir et al. [10,11] across the multilevel borehole network (S1 Table). Water was sampled using a bladder pump (flow rate of 100 ml min -1 ) with a Teflon outlet tube (diameter: 0.6 cm) (Geotech Environmental Equipment, Inc., USA) following a low-flow micro-purging protocol [22]. This pump minimises sample mixing and degassing [10]. Otherwise a peristaltic pump and 20 ml syringe connected to a Teflon tube (diameter: 0.5 cm) with 3-way stopcock was used. Triplicate 50 ml surface water samples from pipes or open drains were collected manually in HDPE screw top bottles and stored at 4˚C until analysed. One replicate was filtered in the field (0.45 μm cellulose acetate filter). As per historical data an electronic dipper and In-Situ Multi-parameter Probe were used in the field.

Water and gas analyses
Water samples (both current and historical) were analysed within two weeks of collection for the following:

Stable isotope analysis
In Sep 2014 and June 2017 groundwater samples (40 ml) were collected on two dates, filtered in the field through 0.2 μm polyethersulfone filters (Sartorius Stedim Biotech GmbH, Germany) and stored at -20˚C in 50 ml polyethylene screw cap tubes.

Statistical analysis
Different methods (t-test, oneway ANOVA and Tukey's HSD test (IBM SPSS Statistics version 24)) were used to determine possible correlations and differences between nutrient, isotopic, and gaseous data and other hydrochemical parameters within the management groups.

Partial N balance
Annual total N-inputs were estimated 364 kg N ha -1 (296 kg N ha -1 -fertilizers, 68 kg N ha -1feed concentrates) whereas national average for intensive farms is 223 kg N ha -1 ( [17]) (Table 1). Irrigation N was composed of slurry and DSW and it was accounted as organic fertilisers. In this area, between 2013-2014, the annual sum of wet and dry deposition was 6 and 8 kg ha -1 per year (2-4 kg ha -1 per year of NH 4 + for wet and dry deposition, respectively) [27].
Mineralization is a larger N source than ammonia deposition but variable and dependant on soil and climatic conditions. Murphy et al. [28] found that on sandy loam soils, with 6 kg ha -1 N deposition, net N mineralisation rates of 156 kg ha -1 per year occurred while of 161 kg ha -1 per year on clay loam. With an N-output of 140 kg N ha -1 exported in milk and slurry, the estimated N-surplus is 224 kg N ha -1 , which is stored in soil, used by the plant, lost from the system through leaching and runoff, or lost in the atmosphere as N 2 , N 2 O or as gaseous ammonia [29]. Ammonia emissions vary depending on climatic conditions during N application and applied N source; Misselbrook et al. [30] calculated that on English soils the emission factor from slurry and DSW is 34% of total ammonia N applied and 6% of total ammonia N applied, respectively.

Dissolved-N 2 O and excess-N 2
Dissolved gas (dissolved-N 2 O and excess-N 2 ) results varied spatially and with depth. In water samples, the excess-N 2 concentration averaged 2.28 mg N 2 -N l -1 , with higher values at the interface and bedrock layer, while dissolved N 2 O was highest in the subsoil i.e. 0.024 mg N 2 O-N l -1 [15].  An integrated nitrogen assessment to inform management on intensive dairy systems In Sep 2014, the average farm excess-N 2 was 1.90 mg N 2 -N l -1 (max. 82 mg N 2 -N l -1 ; min. 0.0004 mg N 2 -N l -1 ) and the average drainage system was 0.076 mg N 2 -N l -1 (max, 0.18 mg N 2 -N l -1 ; min, 0.006 mg N 2 -N l -1 (S2 Fig, S2 Table). The up-gradient unit was characterised by emissions of 0.54 mg N 2 -N l -1 . In the down-gradient area at shallow depth, excess-N 2 was elevated in well-to moderately-drained soils, with peak concentrations of 3.36 mg N 2 -N l -1 at 2.95-4.5 m bgl and 4.32 mg N 2 -N l -1 at 4.5-6 m bgl. Up-gradient imperfectly drained soils had elevated concentrations (4.12 mg N 2 -N l -1 ) at 6-9 m bgl and these were higher in central parts of the farm (6.52 mg N 2 -N l -1 and 6.82 mg N 2 -N l -1 ) at 11-13 and 16 m bgl, respectively (S2 A wide range of dissolved N 2 O vs. total emissions (N 2 O + N 2 ) was found, which suggests a variable rate of denitrification. If the denitrification rate is high keeping the NO 3 --N concentration below the contamination threshold, excess-N 2 is released due to completion of the process. Conversely, high dissolved-N 2 O occurs where denitrification is limited [3]. Nitrification contributes to N 2 O production with stable isotopes (δ 18

Investigating connectivity
Stable isotope compositions for all water samples showed low spatial variability, with values between -6.4 and -3.4‰ for H 2 O-δ 18 (Fig 4). To examine the groundwater interaction with the drainage system, water stable isotope values were compared with the Global Meteoric Water Line (GMWL, δD = 8 � δ 18 O+10) [38] to infer their composition and provenance (Fig 4). Farm values had a lower slope than the GMWL. Water samples (from screened intervals of piezometers and boreholes) had stable isotopic values close to the GMWL. Some sub-surface locations had the same water isotope values as many groundwater boreholes and piezometers. However, for other locations, e.g. along the surface open drain, there was an enrichment in δ 18 O but not δD relative to the GMWL in 2014, while in both δ 18 O and δD in 2017 (Fig 4), therefore showing a second signature due to evaporation in the open drain. A third signature from the lake system exhibited enrichment in both δ 18 O and δD-H 2 O (Fig 4). Groundwater flow direction is from north to south, mirroring topography [14].

Spatial variation in nitrate δ 18 O and δ 15 N in water samples
In 2014, the average δ 15 N composition of NO 3 was 14.2‰ (6.2-54.9‰, Fig 5,   An integrated nitrogen assessment to inform management on intensive dairy systems Importantly, comparing these values with [14] for the same wells, 75% of δ 15 N-NO 3 and  [14] that there is a consistent temporal signal. In 2014, NO 3 --N was spatially variable with some locations exceeding 5.65 mg l -1 (Fig 2, S2   Table). In 2008, three main areas: the north end (surrounding 5, well-drained soil), with a value of 9.5 mg NO 3 --N l -1 (±2.9); the central area (surrounding 14, well-drained soil), with a value of 10.3 mg NO 3 --N l -1 (±4.5); and south end (surrounding 28, well-to moderatelydrained soil), with a value of 7.3 mg NO 3 --N l -1 (±2.4) [14] were elevated. High NO 3 --N was attributed to old DSW irrigation areas and farmyard leachate (see [14]). In Sep 2014 NO 3 --N < 5.65 was evident for all locations (Fig 5). In terms of N source groundwater NO 3 --N isotopic composition clustered within the manure/sewage value range and along a 1:1-1:2 slope, suggesting a common organic source for these, and denitrification as the main biotransformation process (Fig 5) 5). The subset of data points, close to the 1:1 isotope ratio line, where denitrification was the dominant process, was further examined by eliminating points with a fertiliser and ammonia volatilisation signature (Fig 5). A plot of δ 18 O-NO 3 versus NO 3 --N revealed three main groups, each identifying a specific condition of inputs and extent of denitrification (Fig 5). The first group had a low NO 3 --N concentration (<5.65 mg NO 3 --N l -1 ) associated with limited denitrification isotopic signature (<10‰) and low inputs. The second group had a low NO 3 --N concentration but high denitrification isotope signature (>10‰), indicating that excess inputs of NO 3 --N have been denitrified. The final group had high NO 3 --N (>5.65 mg NO 3 --N l -1 ) and low denitrification isotopic signature (<10‰), highlighting contamination but limited denitrification. The first group was discarded from analysis as not influential in terms of denitrification and contamination, while groups with NO 3 --N above 5.65 mg NO 3 --N l -1 or δ 18 O-NO 3 above 10‰ were selected. Poorly-drained soils are characterised by high denitrification potential and low NO 3 --N, whereas well-drained soils had low denitrification potential and higher NO 3 --N. These isotopic enrichment patterns followed soil permeability patterns which were consistent with the findings of Fenton et al. [13] (Fig 5).
Natural -N polluted wells occur mainly in the central area of the farm (Fig  3) with moderate-to imperfectly-drained soil. A high NH 4 + -N concentration was evident in wells showing relative enrichment in δ 18 O-NO 3 and with a synthetic fertilizer source signature ( Fig 5). However, a high NH 4 + -N concentration could occur without this signature, due to in-situ transformation of NO 3 to NH 4 + by DNRA [11,15].

Conceptual diagram of the site to inform soil nitrogen attenuation potential
Integrating all data (Fig 6), four main location groups were distinguished with respect to potential for soil N attenuation (Fig 1, S2 Table): Group 1 (high N attenuation potential): locations with NH 4 + -N and NO 3 --N concentrations below MAC. Drainage class of these soils are imperfectly to moderately-well drained and drainage water from these areas does not represent a threat to water quality. This condition reflects relatively high NO 3 --N attenuation rates, via denitrification. Generally low N 2 O emissions inferred completeness of denitrification, although some locally elevated N 2 O production suggested concurrent secondary processes (e.g. nitrification), with a possible threat to air quality. Group 2 (moderate N attenuation potential): locations with a NO 3 --N concentration below the contamination threshold, but with high N 2 O production. Drainage classes are well to moderately-drained and drainage water from these soils does not represent a threat to water quality, but instead is a problem in terms of GHG emissions. Such emissions are due to a low capacity for denitrification, resulting in low NO 3 --N, possibly coupled with additional N 2 O emissions from nitrification. Group 3a (low N attenuation potential, DSW irrigation): locations with a NO 3 --N concentration above significant contamination. Drainage classes are mainly well to moderately-well drained, with a higher permeability than those in Group 1 and drainage water from these locations represents water and air quality issues (due to high N 2 O emissions, which indicate incomplete denitrification). This results from a medium-low potential for NO 3 --N attenuation by denitrification and if drained will present a threat to groundwater and GHG emissions.
Group 3b (low N attenuation potential): locations with low NO 3 --N but NH 4 + -N above NH 4 + -N MAC and high N 2 O emissions. Drainage classes are well to imperfectly-drained with generally low potential for denitrification and drainage water is a threat to groundwater and GHG emissions.
De Klein et al. [52] observed that more than 50% of EU dairy farms have high N surpluses (>200 kg N ha -1 ), with only 7% showing values below 100 kg N ha -1 . The N surplus on our Irish farm was high and above the national average of 175 kg N ha -1 [53]. As a direct continuum has been shown to exist N can transported and transformed as follows (Fig 6): 1. a shallow migration pathway in poorly-or imperfectly-drained soils with high water purification capability. The drainage system does not disrupt this capacity but instead conveys clean water to the exit point of the farm; 2. a deep migration pathway in moderately-and well-drained soils where the water purification capability is lower. This facilitates leaching of N, which is then converted at depth to NH 4 + -N and migrates off site along deeper groundwater pathways. An integrated nitrogen assessment to inform management on intensive dairy systems

Development of farm management strategy for nitrogen
The farm is operating within the current regulations [54-56], following the EU Nitrates Directive [57], but these assume homogeneous soil-subsoil-geology without variation in N natural attenuation capacity. The present study shows that four attenuation capacity areas exist, with Group 3a and 3b areas representing the most vulnerable. N loads need to be diverted away from these areas e.g. DSW produced needs to be reduced and irrigation on the farm should avoid group 3a and b areas, and be reduced on group 2 areas.
Historically, the DSW irrigation system has been moved around (e.g. after a long period in 2005 it was moved from heavier soils along the central drainage ditch to avoid overland flow losses) and since then it is positioned behind the farmyard on well-to moderately-drained soils. This static position has concentrated a high N load in this area and can be correlated with elevated NO 3 --N in down-gradient groundwater. Presently all DSW is collected in a three-chamber separation farmyard underground tank (total capacity~55,000 l) with a variable retention time depending on volume of washings and rainfall. In periods where land is saturated and spreading with Roto-Rainer is not possible, DSW can be stored in an earthen bank lagoon. This DSW is spread with an umbilical cord system within the downstream farm on the three main sections below the built-up area in Fig 1 (plots with locations 28, 33 and 35).
Farmyard measures must be optimised to manage the volume of DSW produced on the farm, through 1) Reduction of DSW production-Diversion of all clean water to a clean water outfall, preventing clean water from becoming soiled and restructuring of the yard to minimise DSW generation; 2) Optimisation of DSW spreading-Use of the information gained from the present study (heterogeneous soil functionalities) to better use the entire farm to spread DSW. The following management options are considered to be viable in terms of reducing the N load: 1. Short term management options: 1) Moving DSW irrigation to high N attenuation areas (Group 1 imperfectly-to moderately-well drained soil locations), or moderate N attenuation areas (Group 2 well-to moderately-drained soil locations), where no overland flow component could affect the open ditch network; 2) Removal of the N load from vulnerable areas that will allow the N biogeochemical time lag to diminish over time, which should result in a decreased groundwater NO 3 --N concentration.
2. Long term management option: minimise the N source whilst using the lagoon for storage. This provides flexibility for DSW to be 1) land spread when its fertilizer replacement value is high, or 2) discharged to a constructed wetland installed at the end of the farm (Group 1 location) at other times, e.g. over winter. The lagoon concept has been considered on dairy farms [58,59]. The loss of land for the lagoon would not be severe as full utilisation of the area currently being used for DSW application is difficult at times due to ground trafficability and grass palatability. In addition, the labour lost through the management of an irrigation system would be gained for other activities on the farm. Significant energy saving is also achievable with the significant reduction in pumping of DSW. One consideration here is to design the system for expansion in the event that the discharge licence is not met over time, in which case a de-sludging specification should be considered.

Conclusions
Integrating surface and sub-surface spatial and temporal data on an intensive dairy farm provided knowledge on N load and surplus, N source-transformation and fate along surface and subsurface pathways. This information enabled the dairy farm to be divided into four distinct areas based on N attenuation capacity. This method has the ability to match future N loads with attenuation capacity, thereby optimising the farm system in terms of environmental sustainability.