Products considered

Twelve agricultural food categories were considered, divided into six categories of vegetable products and six of livestock products (Table 1). These food categories cover virtually all livestock production in Europe, except fish and seafood products, which are not simulated in CAPRI or MITERRA. Also most vegetable food products in the EU27 were considered, excluding production of wine and hops for beer, as well as non-food crop production, such as fibre crops. Overall, the analysis covered 0·97 of EU27 food crop production.

Table 1.

Food categories considered and corresponding crops or products

Nitrogen indicators

N footprint

The N footprint (Φ) was used as an indicator of the total direct N-losses to the environment that occur for the production of one unit of (food) product, measured in g N/kg food product. In accordance with Leach et al. (Reference Leach, Galloway, Bleeker, Erisman, Kohn and Kitzes2012) the N footprint was calculated as total N emission intensity for one unit of product. Nitrogen emissions were calculated based on a soil N-budget approach for vegetable food products, or an ‘extended soil budget’ approach for livestock products, referred to in the following as a Nitrogen Footprint Budget (NB^Φ). In contrast to the farm N-budget (NB^F) (Leip et al. Reference Leip, Britz, Weiss and de Vries2011b ), NB^Φ uses feed imports to the ‘farm’ as throughput flow and ‘replaces’ it in the calculation with the specific soil N-budget of the feed product.

As such, the N footprint budget NB^Φ was based on a partial life-cycle assessment (LCA), however, without considering other Nr losses such as NO_x emissions from energy use or N₂O emissions from land use change, which are considered in a full cradle-to-gate LCA. The quantification of the N footprint in the current paper did not include wastage that might occur during further processing, retail or preparation of the food products, which are considered in the N footprint of Leach et al. (Reference Leach, Galloway, Bleeker, Erisman, Kohn and Kitzes2012). Thus, a ‘farm-gate’ Nitrogen Footprint Φ^G was quantified, in contrast to the ‘farm-to-fork’ Nitrogen Footprint Φ^F as estimated by Leach et al. (Reference Leach, Galloway, Bleeker, Erisman, Kohn and Kitzes2012).

As product units, the biomass of the product at farm gate $m_{\,product}^G $ was used to calculate φ^G [kg N/kg product].

(1) $$\phi ^G = \displaystyle{{N_{in}^S - N_{out}^G} \over {m_{\,product}^G}} $$

where $N_{in}^S $ is the sum of the input N flows according the definition of a soil N budget and $N_{out}^G $ is the sum of the useful outputs generated.

Input flows were the application of fertilizers, biological N fixation and atmospheric deposition. Fertilizers included both mineral and organic fertilizers, considering only external inputs and discounting throughput flows from both input and output sides (e.g. manure). The main useful output was the product for which the N footprint was being calculated (see earlier), but ‘waste’ streams N _waste that are recycled and used as input for other processes were included as well. These flows include N accumulating in the soil – soil stock changes (ssc) – that are available for future crop growth (for a discussion of accounting of soil stock changes see Leip et al. Reference Leip, Britz, Weiss and de Vries2011b ), crop residues (cr) used to fertilize other fields or used as feed, manure (man) used to fertilize other fields and ‘pre-gate’ food processing wastes (waste) recycled in a food processing chain or used in the production of other durable goods. The latter concerned mainly slaughter house wastes (which were considered to be within the ‘farm gate’ boundaries), modelled here as the difference between animal live weight and animal carcass. The recycled N-flows were expressed as a fraction of the total waste flows that were recycled as given in Eqn (2).

(2) $$N_{out}^G = N_{\,product}^G + N_{rec}^G = N_{\,product}^G + \mathop \sum \limits_{{ {x = ssc,cr,} \atop {man,waste} \cr}} \{ f_{x,rec} \; \times N_{x,waste\;} \} $$

For the current study, the fractions f _x,rec of recycled N streams were set to zero for surplus-manure, since – even if applied to another crop – it is not necessarily needed in a situation of general over-supply of N for European agriculture. Crop residues are beneficial for soil fertility (McIntyre et al. Reference McIntyre, Herren, Wakhungu and Watsn2009) and were therefore considered completely as useful output with a f _cr,rec of one. A shift of crop residues between crop activities occurs mainly through the application of solid manure (bedding material) and following crop rotation. Also for increases in soil N stocks f _ssc,rec =1 was assumed. In CAPRI and MITERRA, soil stock changes were not quantified but may result in the case of $N_{out}^S \gt N_{in}^S $ , in which case soil depletion occurs.

Nitrogen investment factor

In addition, a dimensionless Nitrogen investment factor was calculated at the farm gate (NIF^G ) which measures the total (external) N required to produce one unit of product in terms of N contained. Thus, in contrast to the N footprint which is expressed per unit biomass of product, the N investment factor is expressed per unit N of product, facilitating the comparison of different products. Also, in contrast to the N footprint, the N investment factor is an index for total N use, including both N losses and N recovered in the product.

(3) $$NIF^G = \displaystyle{{N_{in}^S - N_{rec}^G} \over {N_{\,product}^G}} = \phi ^G \times \left( {\displaystyle{{N_{\,product}^G} \over {m_{\,product}^G}}} \right)^{ - 1} + 1$$

The N investment factor for sugar and oils (which do not have significant quantities of N) were calculated based on the N:biomass ratio in the primary crop (thus sugar beet and oilseeds). The N contained in sugar beet and oilseeds ends up in by-products as molasses and oilseed meals, which are usually used as feed.

Models

Calculations were carried out with the CAPRI and MITERRA models. The reference year was 2004, which was the available base year of CAPRI. All statistical input data were based on 3-year averages of the period 2003–2005. Data from EU25 are presented, covering all the countries of EU27 apart from Malta and Cyprus, for which the quality of data was insufficient. The N footprint and the N investment factor were estimated for the supply of food products, whereby a weighted average is calculated between regional domestic production and imported products.

CAPRI

The CAPRI model (Britz & Witzke Reference Britz and Witzke2012) is an agricultural policy impact assessment tool focusing on the European Union. Apart from the simulation model, the CAPRI system includes a complete and consistent modelling database, combining information from official and harmonized data sources such as EUROSTAT, FAOSTAT and OECD. For the EU27, the database has a spatial resolution of 225 regions. Data gaps are closed and data inconsistencies are removed by changing affected data according to their estimated uncertainty. The resulting database is complete and consistent with respect to agricultural crop and livestock production activities as well as the required inputs (farm inputs, costs) and outputs (production, return etc.) and can be used for the quantification of a series of agri-environmental indicators, among others all relevant N flows in the agricultural sector of EU countries.

Livestock feed was calculated in a feed distribution tool of CAPRI using feed data aggregates from market balances on a country level, regional fodder availability and animal requirements. Sources of N to soils are atmospheric deposition, biological N-fixation, and returns from agricultural production systems (crop residues, manure, soil N stock changes) as described in Leip et al. (Reference Leip, Britz, Weiss and de Vries2011b ). Briefly, N-deposition was from the European-scale EMEP MSC-W Chemical transport model (Simpson et al. Reference Simpson, Fagerli, Jonson, Tsyro, Wind and Tuovinen2003, Reference Simpson, Aas, Bartnicki, Berge, Bleeker, Cuvelier, Dentener, Dore, Erisman, Fagerli, Flechard, Hertel, van Jaarsveld, Jenkin, Schaap, Shamsudheen Semeena, Thunis, Vautard, Vieno, Sutton, Howard, Erisman, Billen, Bleeker, van Grinsven, Grennfelt and Grizzetti2011). N-fixation was based on fixed coefficients for pulses and grass, taking into consideration the share of clover in grasslands according to data from the Farm Structure Survey. National mineral fertilizer data were from the International Fertilizer Association (IFA; www.fertilizer.org/ifa/ifadata/search) and the European Fertiliser Manufacturers Organization (EFMA). Manure excretion was calculated with an animal budget accounting for N in feed, N in products and N-retention. Fertilizer was distributed to crops according to crop needs and N availability in N-sources and factors on manure and crop residues availability, minimum share of N obtained from mineral sources and ‘luxury’ factors accounting for expected N losses and farmer's ‘security margin’ (Britz & Witzke Reference Britz and Witzke2012).

An LCA approach, implemented in CAPRI for greenhouse gas accounting (Leip et al. Reference Leip, Weiss, Wassenaar, Perez-Dominguez, Fellmann, Loudjani, Tubiello, Grandgirard, Monni and Biala2010; Weiss & Leip Reference Weiss and Leip2012), was adapted slightly for calculation of the N footprint. The following feed and forage categories were included in CAPRI: feed cereals, protein rich feeds (e.g. soybean meal), energy rich feeds (e.g. cassava meal, sugar beet molasses), maize and grass forages from arable land and grass from permanent grassland, straw, feed arising from dairy products (e.g. whey, milk) and by-product feed (e.g. citrus pulp). Nitrogen-flows from feed cultivation were allocated to livestock products in three steps. First, for some processed feedstuffs (such as soybean cake and soybean oils, etc.), flows were allocated from primary product to the feedstuff based on mass weight. Then all the N-flows from feed were allocated to the animals ingesting the feed according to the ratios estimated in the feed distribution tool. As part of the feed is imported from countries outside the EU, for which a full soil N-budget cannot be calculated, EU-averages were used as a default for imported crop products and feed concentrates. Finally, N-flows from animals were allocated to animal products using the outputs per animal production activity and the N contents of the animal products.

MITERRA

MITERRA-Europe is an environmental impact assessment model, which calculates emissions of N and greenhouse gases on a deterministic and annual basis using emission and leaching factors (Velthof et al. Reference Velthof, Oudendag, Witzke, Asman, Klimont and Oenema2009; Lesschen et al. Reference Lesschen, van den Berg, Westhoek, Witzke and Oenema2011). The model was developed to assess effects and interactions of policies and measures in agriculture on N losses at a regional level (NUTS2) in the EU27 (Velthof et al. Reference Velthof, Oudendag, Witzke, Asman, Klimont and Oenema2009). MITERRA-Europe is partly based on the models CAPRI (Britz & Witzke Reference Britz and Witzke2012) and GAINS (Klimont & Brink Reference Klimont and Brink2004), supplemented with an N leaching module, a soil carbon module and a mitigation module for greenhouse gas and NH₃ emissions and NO₃ leaching measures. Input data consists of activity data (e.g. livestock numbers, crop areas), spatial environmental data (e.g. soil and climate data) and emission factors (IPCC and GAINS).

The main input data for MITERRA-Europe were crop areas, animal numbers and feed use at the regional level. Crop areas and feed use were taken directly from CAPRI (see earlier section on CAPRI). The N content of each feed type was assumed to be the same among EU countries. Country-specific N contents were used only for grass (Velthof et al. Reference Velthof, Oudendag, Witzke, Asman, Klimont and Oenema2009). Animal numbers were from GAINS at a national level, and distributed over the NUTS2 regions according to CAPRI livestock data. Data on primary animal and crop production and annual N fertilizer consumption were collected from FAOSTAT at national level. Since animals are not always slaughtered in the same country where they are raised, corrections were made for the export and import of live animals as described in Lesschen et al. (Reference Lesschen, van den Berg, Westhoek, Witzke and Oenema2011).

For the calculation of N input and N emissions in terms of animal products, the feed consumption was allocated to crop areas as described in Lesschen et al. (Reference Lesschen, van den Berg, Westhoek, Witzke and Oenema2011). For soybean, which is the main feed product from outside the EU, representative N inputs and N emissions from Brazil were used, based on Smaling et al. (Reference Smaling, Roscoe, Lesschen, Bouwman and Comunello2008).

Country-specific N excretion rates of livestock were obtained from the GAINS model (Klimont & Brink Reference Klimont and Brink2004). The total manure N production was calculated at the NUTS2 level using the number of animals and the N excretion per animal, then corrected for N losses in housing and storage. Manure was distributed over arable crops and grassland according to Velthof et al. (Reference Velthof, Oudendag, Witzke, Asman, Klimont and Oenema2009), taking into account the maximum manure application of 170 kg N/ha, or higher in the case of a derogation from the Nitrates Directive. Mineral N fertilizer was distributed over crops relative to their N demand, taking account of the amount of applied manure and grazing manure and their respective fertilizer equivalents. The N demand was calculated as the total N content of the crop (harvested part plus crop residue), multiplied by a crop-specific uptake factor, set at 1·0 for grass and 1·1 and 1·25 for cereals and other arable crops respectively (Velthof et al. Reference Velthof, Oudendag, Witzke, Asman, Klimont and Oenema2009). Further N inputs include biological N fixation, which was estimated as a function of land use and crop type (legumes), and N deposition that was derived at NUTS2 level from the European-scale EMEP MSC-W Chemical transport model (Simpson et al. Reference Simpson, Aas, Bartnicki, Berge, Bleeker, Cuvelier, Dentener, Dore, Erisman, Fagerli, Flechard, Hertel, van Jaarsveld, Jenkin, Schaap, Shamsudheen Semeena, Thunis, Vautard, Vieno, Sutton, Howard, Erisman, Billen, Bleeker, van Grinsven, Grennfelt and Grizzetti2011).

Nitrogen flows

Nitrogen (NH₃, N₂O, N₂, NO_x, N-leaching and runoff) emissions were calculated following a mass-flow approach in both CAPRI and MITERRA models, based on the MITERRA-Europe model (Velthof et al. Reference Velthof, Oudendag, Witzke, Asman, Klimont and Oenema2009). It represents the N cycle in agricultural systems (de Vries et al. Reference De Vries, Leip, Reinds, Kros, Lesschen and Bouwman2011; Leip et al. Reference Leip, Britz, Weiss and de Vries2011b , Reference Leip, Weiss, Britz and Flichman c ). Any N-emissions occurring in earlier stages, such as during the storage of manure, were subtracted from the N pool before calculating emissions at a later stage, such as following the application of manure.

Emissions of NH₃ from livestock manure take place during housing and manure storage, after application to the soil, and from grazed land. Country-specific emission factors and estimates of the efficiency of ammonia abatement measures were taken from the GAINS model (Klimont & Brink Reference Klimont and Brink2004). Emissions of N₂O from agriculture consist of emissions from manure management, soil emissions from the application of mineral fertilizer and animal manure, crop residues, grazing, and indirect emissions from N lost due to leaching and runoff, and from volatilized and re-deposited N. All N₂O emissions were calculated using emission factors from the IPCC (Reference Eggleston, Buendia, Miwa, Ngara and Tanabe2006) guidelines. The emission factor for NO_x was derived from Skiba et al. (Reference Skiba, Fowler and Smith1997) and was set at 0·3% of the N input. In MITERRA, N leaching was calculated by multiplying the soil N surplus by a region-specific leaching fraction, based on soil texture, land use, precipitation surplus, soil organic carbon content, temperature and rooting depth. Surface runoff fractions were calculated based on slope, land use, precipitation surplus, soil texture and soil depth (Velthof et al. Reference Velthof, Oudendag, Witzke, Asman, Klimont and Oenema2009).

In CAPRI, leaching rates including surface runoff were taken from IPCC (Reference Eggleston, Buendia, Miwa, Ngara and Tanabe2006) but related to the N soil surplus instead of N application. Therefore, leaching is restricted to the level of the N soil surplus corrected by a minimum denitrification rate, which is assumed to be 2·5 times the N₂O emissions. The share of agricultural land affected by leaching is derived via the land where access supply of water (rainfall in rainy season minus potential evapotranspiration) is larger than the water holding capacity of the soils, not taking into account irrigation. Data sources applied are 5×5 maps on long-term average of monthly rainfall (Hijmans et al. Reference Hijmans, Cameron, Parra, Jones and Jarvis2005), long-term average of monthly potential evapo-transpiration (PET) of the reference land use ‘grassland’ (R. Hiederer, 2010, personal communication, based on data from Hijmans et al. Reference Hijmans, Cameron, Parra, Jones and Jarvis2005) and soil water holding capacity provided along with the ISRIC-WISE soil properties data set (Batjes Reference Batjes2006).