Greenhouse gas emissions from the EU livestock sector: A life cycle assessment carried out with the CAPRI model Academic research paper on "Earth and related environmental sciences"

ELSEVIER

Contents lists available at SciVerse ScienceDirect

Agriculture, Ecosystems and Environment

journal homepage www.elsevier.com/locate/agee

Greenhouse gas emissions from the EU livestock sector: A life cycle assessment carried out with the CAPRI model

Franz Weiss, Adrian Leip *

Joint Research Centre, Institute for Environment and Sustainability (IES), Climate Change and Air Quality Unit, Via E. Fermi 2749,21027 Ispra, VA, Italy

ARTICLE INFO

ABSTRACT

Article history:

Received 2 September 2011

Received in revised form 2 December 2011

Accepted 21 December 2011

Available online 28 January 2012

Keywords:

Life cycle assessment Livestock Greenhouse gases Agriculture Land use change

This study presents detailed product-based net emissions of main livestock products (meat, milk and eggs) at national level for the whole EU-27 according to a cradle-to-gate life-cycle assessment, including emissions from land use and land use change (LULUC). Calculations were done with the CAPRI model and the covered gases are CH4, N2O and CO2. Total GHG fluxes of European livestock production amount to 623-852 MtCO2-equiv., 182-238 MtCO2-equiv. (28-29%) are from beef production, 184-240 MtCO2-equiv. (28-30%) from cow milk production and 153-226 Mt CO2-equiv. (25-27%) from pork production. According to IPCC classifications, 38-52% of total net emissions are created in the agricultural sector, 17-24% in the energy and industrial sectors. 12-16% Mt CO2-equjv. are related to land use (CO2 fluxes from cultivation of organic soils and reduced carbon sequestration compared to natural grassland) and 9-33% to land use change, mainly due to feed imports. The results suggest that for effective reduction of GHG emissions from livestock production, fluxes occurring outside the agricultural sector need to be taken into account. Reduction targets should address both the production side as defined by IPCC sectors and the consumption side. An LCA assessment as presented here could be a basis for such efforts.

© 2012 Elsevier B.V. All rights reserved.

1. Introduction

The FAO report "Livestock long shadow: environmental issues and options" (2006) finds that livestock production is a major contributor to the world's environmental problems (Galloway et al., 2010; Steinfeld and Wassenaar, 2007; Steinfeld et al., 2010; Sutton et al., 2011), contributing about 18% to global anthropogenic greenhouse gas (GHG) emissions, although highly variable across the world (FAO, 2006). It is based on a food-chain approach, bringing into light also contributions normally 'hidden' in other sectors when the internationally agreed methodology of GHG emissions accounting within the United Nations Framework Convention on Climate Change (UNFCCC) is used. For Europe only few specific assessments on the climate impact of livestock production systems are available which cover a wide range of products and the whole of the EU-27. For example, Lesschen et al. (2011) calculated emission profiles for animal products of European countries and FAO (2010) has differentiated the GHG impact of dairy systems by world regions. Several studies apply detailed life cycle assessments (LCAs) of GHG emissions for specific areas and animal products, however using different approaches, scopes and functional units, making them hardly comparable. Emissions from induced land use change (LUC)

* Corresponding author. Tel.: +39 0332 786327; fax: +39 0332 785022. E-mail address: adrian.leip@jrc.ec.europa.eu (A. Leip).

0167-8809/$ - see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.agee.2011.12.015

are usually not considered (de Vries and de Boer, 2010; Garnett, 2009).

The UNFCC accounting system (IPCC, 1997, 2003, 2006) is strongly tied to the needs of current international emission reduction agreements and its purpose is to report GHG fluxes in those countries where they are created and where they can be avoided by appropriate national measures. Therefore, according to UNFCC rules, GHG emissions are assigned to a specific country and a specific sector if they are directly created by a production or consumption process in the respective country and the respective sector. In contrast, fluxes of inputs and intermediate products, which have been generated in other countries or sectors, are not linked to their final use. This is particularly important for emissions and removals from land use and land use change (LULUC), which according to UNFCC are reported in a chapter separated from the economic activities driving the processes behind. Although the overwhelming part of land use changes is driven by agricultural or forestry activities, they are not assigned to those sectors. As a consequence, the UNFCC accounting logic leads to an incentive to transfer emission intensive production processes to countries with no emission reduction obligations and then importing the respective products to industrialized countries. This may even lead to increasing emissions if the environmental standards in the countries without obligations are lower and due to emissions from additional transport.

The concept of life cycle assessment attempts to overcome this shortcoming by assigning all fluxes created directly or indirectly by

the production, consumption and disposal to the final product. This requires not only an extension of the sectoral scope but also of the regional scope, since imported inputs have to be considered too. We are far from a global data base which could serve as a basis for such an accounting system, but research groups have made considerable progress in the last decade to improve available information for life cycle assessments which could serve as basis for evaluating the global warming effect of certain products. The current study is the first attempt to give such a comprehensive picture for all EU countries and the most important animal products.

In this paper we present a detailed cradle to farm gate attribu-tional life cycle assessment of net GHG fluxes from the livestock sector for the year 2004 and all EU-27 countries carried out with the CAPRI modeling system (Britz and Witzke, 2008). We consider on-farm and off-farm fluxes related to livestock rearing and the production of feed, including fluxes from the production of mineral fertilizers, pesticides, energy, feed transport and feed processing, as well as emissions from LULUC.

2. Methodology

Emissions and removals of GHGs from the EU livestock sector were calculated for methane (CH4), nitrous oxide (N2O) and carbon dioxide (CO2), for 21 different emission sources, 7 animal products (beef, cow milk, pork, sheep and goat meat and milk, eggs and poultry meat), 218 European (so-called NUTS 2) regions, and 26 member states (Belgium and Luxemburg are treated together).

Calculation was done with the CAPRI model (see Britz and Witzke, 2008; Tukker et al., 2011; Jansson and Heckelei, 2011) for the base year 2004. CAPRI is both a data base and a simulation model for the agricultural sector of the EU. The CAPRI modeling system uses a unified, complete and consistent data base, which is derived from various sources such as national statistics on slaughtering, herd size, crop production, land use, farm and market balances and foreign trade as well as regional statistics on the same issues from the REGIO database,1 if available. Moreover, it integrates economic, physical and environmental information in a consistent way (see, e.g. Leip et al., 2011a, 2011b; Britz et al., 2011).

CAPRI estimates GHG fluxes for all emission sources of the agricultural sector as defined by the IPCC (1997). For details on the estimation of fluxes of nitrogen and GHGs, see Leip et al. (2010) and Perez-Dominguez et al. (2009). Generally, the quantification of GHG emissions follows the IPCC (2006) guidelines. The quantification of methane emissions from enteric fermentation and manure management follows a Tier 2 approach for cattle activities and a Tier 1 approach for swine, poultry, sheep and goats. Feed digestibility is calculated endogenously on the basis of the feed ration. Nitrogen fluxes are calculated according to a mass flow approach developed for the MITERRA-EUROPE model using data of the GAINS database (Velthof et al., 2009; Klimont and Brink, 2004). CO2 emissions from on-farm energy usage are based on Kränzlein (2009), CO2 fluxes from carbon sequestration of grassland and cropland are estimated on the basis of data from Soussana et al. (2007, 2009) as described in Leip et al. (2010). For the calculation of emissions from land use change (LUC) see Section 2.1 and Annex.

Emissions, in a first step, are generally calculated per hectare of land or per animal head and thus consider only emissions and removals which are directly created by the respective agricultural activity, like, i.e. the fattening of young bulls, in the respective country or region. Product-based GHG fluxes (or GHG intensities) are derived from the activity-based emission flux rates in a second step using defined allocation rules. Some emission sources are

calculated directly on a per-product basis, such as CO2 emissions from feed transport and GHG emissions from LUC. Finally, emission intensities are carried through the supply chain and allocated to the final functional units, again following defined allocation rules. For emissions associated to imported feed products from outside Europe, we use the average European values, as the information required to calculate country-specific emission values globally was not available. Exceptions are emissions from LUC and the cultivation of organic soils, which have been calculated for all world regions.

Table 1 provides a complete list of emission sources considered in the life cycle assessment and the associated gaseous fluxes. Moreover, it indicates whether the fluxes are caused directly by livestock rearing activities or cropping activities for the production of feed, and the economic sector the emissions are generated.

2.1. Calculation of emissions from land use change

For LUC, we consider CO2 fluxes from carbon stock changes in below and above ground biomass and dead organic matter, CO2 fluxes from soil carbon stock changes, and CH4 and N2O fluxes from biomass burning. Carbon factors are generally taken from the IPCC 2006 guidelines, applying a Tier 1 approach, while data on land use, soils and climate zones are based on the Global Land cover 2000 map, the Harmonized Soil Database (HWSD) and a climate map produced by the JRC (see Carré et al., 2010). The increase of cropland, supposed to be on the cost of other land uses, was derived from the FAO agricultural statistics for the period 1999-2008. In contrast, increases in grassland areas were neglected, because they were assumed to be of minor importance in Europe and outside the scope of the study if happening outside Europe. Since consistent information on transition probabilities (the share of other land uses transformed to cropland) were not available we decided to use three scenarios instead, spanning the space from minimum to maximum fluxes. In the following, in Scenario I we assume that all additional cropland comes from grassland and savannas, Scenario II applies a more likely mix of transition probabilities, and Scenario III can be considered as a maximum emission scenario with a high share of converted forests.

The resulting LUC fluxes are then related to the total production for each of the considered countries or country blocks and crop products. Details on the methodology for the calculation of LUC factors are presented in Annex.

2.2. System boundaries

According to Table 1, the system boundaries of this study are the farm gates, including slaughtering. Thus, we consider emissions on the farm and emissions related to the production of inputs, but not emissions from processing, transport, packaging, retail, consumption and waste of the products, which usually make about 10-20% of total emissions (IDF, 2009).

2.3. Functional unit

The main food productive animal species are covered: (i) beef cattle, (ii) dairy cattle, (iii) small ruminants (sheep and goats), (iv) pigs, and (v) poultry. Results are expressed on the level of animal products. The functional unit is one kilogram of carcass meat, raw milk (at 4%/7% fat content for cow and sheep/goat milk, respectively), or eggs (including the shell).

2.4. Allocation rules

1 http://epp.eurostat.ec.europa.eu/portal/page/portal/region_rities/ regionaLstatistics/data/database.

Allocation of emission fluxes to multiple outputs is required (a) in the feed-supply chain, (b) the livestock rearing, and (c) for

Table 1

Emission sources considered in the GGELS project. Emission source

Enteric fermentation Livestock excretions Manure management (housing and storage)

Depositions by grazing animals Manure application to agricultural soils

Indirect emissions, indirect emissions following N-deposition of volatilized NH3 /NOx from agricultural soils and leaching/run-off of nitrate Use of fertilizers for production of crops dedicated to animal feeding crops (directly or as blends or feed concentrates, including imported feed) Manufacturing of fertilizers

Use of fertilizers, direct emissions from agricultural soils and indirect emissions Use of fertilizers, indirect emissions following N-deposition of volatilized NH3 /NOx from agricultural soils and leaching/run-off of nitrate Cultivation of organic soils

Emissions from crop residues (including leguminous feed crops) Feed transport (including imported feed)

On-farm energy use (diesel fuel and other fuel electricity, indirect energy use by

machinery and buildings) Pesticide use

Feed processing and feed transport

Emissions (or removals) of land use changes induced by livestock activities (feed production)

Carbon stock changes in above and below ground biomass and dead organic matter

Soil carbon stock change Biomass burning

Emissions or removals from pastures, grassland and cropland

Gases Livestock Feed Sector

rearing production

CH4 X Agriculture

N2O X Agriculture

CH4 X Agriculture

N2O X Agriculture

N2O X Agriculture

N2O X Agriculture

CO2 X Energy

N2O X Industry

N2O X Agriculture

N2O X Agriculture

CO2 X LULUC

N2O X Agriculture

N2O X Agriculture

CO2-equiv. X Energy

CO2-equiv. X X Energy

X Energy

CO2 X Energy

CO2 X LULUC

CH4 X LULUC

N2O X LULUC

X LULUC

CO2 X X LULUC

the use of manure. For cases (a) and (b), allocation is done on the basis of the nitrogen content in the products. The N-content serves as an indicator of the causal relationship between product and emissions, protein being the most important nutrient. The only exception is CH4 emissions from dairy cattle (enteric fermentation and manure management), for which energy requirement for lactation and pregnancy is used to allocate emissions to milk and young animals, respectively.

In general the processes for raising and fattening young animals will be allocated to the meat output, while the activities of dairy and suckling cows, sheep and goats for milk or laying hens are split up into the raising of young animals during pregnancy (which is allocated to meat) and the respective product (milk and eggs). The logic behind is, that raising and fattening activities both produce meat by growing animals, even if it will be sold on the market at a later stage. That is, heifers are raised to become dairy cows, but will be slaughtered after having been used as producer of milk and calves for several years. In contrast, the dairy cow activity does not aim at the growth of the cow any more. The main purpose is the production of milk and young calves. So, even if dairy cows are slaughtered and, therefore, deliver meat output, the meat was not created within the dairy cow activity but already before, when the young cow was raised. So, fluxes of the dairy cow activity are allocated to the milk output and the production of young calves.

Finally, we assign emissions from the application of manure entirely to livestock production. However, part of the manure applied on crops is not used for feed thus saving an analogue amount of mineral fertilizer. We account for these emissions with the system expansion approach (see ISO, 2006a, 2006b). The emissions saved are quantified and credited to the livestock product in the respective emission categories (application and production of mineral fertilizers). For a detailed description see Leip et al. (2010).

3. Results

According to our calculations the total GHG fluxes caused by the European Livestock production range between 623 Tg CO2-equiv. and 852TgCO2-equiv., depending on which scenario is used for the calculation of LUC emissions. For Scenario II it amounts to 661 TgCO2-equiv. (see Table 2). The largest share is from beef (28-29%) and cow milk (28-30%) production, followed by pork production (25-27%). All other animal products together do not account for more than (17%) of total fluxes. 323TgCO2-equiv. are created in the agricultural sector, 136TgCO2-equiv. in the energy sector, and 11 Tg CO2-equiv. in the industrial sector (38-52%, 16-22% and 1-2% of total emissions respectively). Correspondingly, emissions from land use are 102TgCO2-equiv. (12-16%), splitting into 76 Tg CO2-equiv. from carbon sequestration and 26 Tg CO2-equiv. from emissions from organic soils. For most animal products, emissions from foregone carbon sequestration dominate enhanced carbon sequestration in managed grasslands leading to net emissions under the position "carbon sequestration". An exception is meat and milk from sheep and goat, where both terms are very close. Emissions from LUC are mainly related to imported feed from non-European countries and range between 54 and 283 Tg CO2-equiv. (9-33% of total emission). In Scenario II, LUC emissions account for 91 TgCO2-equiv. (14% of the Total). Finally, also the contribution of the different animal products to total emissions varies with the scenarios, with the share of pork increasing in scenario III, while the share of beef and cow milk decreases.

Fig. 1 gives more details on emissions accounted for in the LCA analysis in comparison to activity-based calculations according to IPCC emission categories. Emissions from livestock rearing are identical in the activity-based and product-based calculation, while soil emissions include also those from imported feed products. Moreover, the LCA analysis considers also emissions caused

Table 2

Total GHG emissions [Tg CO2-equiv] related to the production of the seven main animal products in EU-27. Data are reported for each greenhouse gas, each UNFCCC sector and the three scenarios (I-III).

Total Agriculture Energy industry Land Use Land use change

CO2-equiv N2O CH4 CO2 N2O Organic soils C-seq CO2-equiv.

i II iii CO2 N2O CO2 i II iii

Beef 179 191 236 45 76 31 3 6 2 6 13 22 69

Cow Milk 184 196 240 38 71 34 3 7 2 17 13 23 69

Pork 153 166 226 35 16 44 3 6 1 32 17 29 90

Poultry Meat 51 55 74 10 0 15 1 2 0 15 6 9 28

Sheep + Goat Meat 23 25 34 5 11 4 0 1 0 0 2 4 14

Eggs 11 12 17 3 6 2 0 0 0 -1 1 2 7

Sheep + Goat Milk 20 21 23 5 0 5 1 1 0 7 1 2 4

Total 623 661 852 140 181 136 11 24 2 76 54 91 283

Fig. 1. Total GHG emissions in EU-27 in 2004, calculated on activity-basis with CAPRI for the IPCC sector agriculture (left column), and based on a cradle-to-gate life-cycle analysis with CAPRI (right column).

by the EU livestock production but created outside the agriculture sector.

On product level the total GHG intensities of ruminants amount to 19-28 kg CO2-equiv. per kg of meat (21-28 kg for beef and 19-28 kg per kg of sheep and goat meat) on EU average, while the production of pork (7-10 kg) and poultry meat (5-7 kg) creates significantly less net emissions. In absolute terms the difference in emissions of pork and poultry meat compared to meat from ruminants is highest for CH4 and N2O emissions, while it is smaller for CO2 fluxes related to feed. Nevertheless both pork and poultry meat production causes lower net emissions also from energy use and LULUC. Compared to other meat products, poultry meat has the lowest emissions. GHG fluxes per kg of cow milk are estimated at 1.3-1.7 kg CO2-equiv. on EU-27 average, those from sheep and goat milk at 2.6-4.1 kgCO2-equiv./kg product. However, data quality in general is less reliable for sheep and goat milk production than for cow milk production, which is important for the assignment of fluxes. The production of eggs leads to the net emission of 2.8-3.2 kg of CO2-equiv. per kg of eggs on EU average. Fig. 2 shows average emission intensities for the seven animal products considered for all EU-27 member states and for the average of EU-27 (see supplementary information for the data by country, gas and animal products). For CO2, emissions from energy, LUC, and land use are shown separately.2

Greenhouse gases emitted for each kilogram of beef, on EU-27 average, are composed of 8.8 kg CO2-equiv. (32-42%) CH4, 5.8kgCO2-equiv. (21-27%) N2O, and 6.5-13.1 kgCO2-equiv. (31-47%) CO2, with approximately equal contributions from the use of energy

2 All data are available at the following address: http://afoludata.jrc.ec.europa. eu/index.php/dataset/files/236.

(3.7kgCO2/kg product) and LULUC (2.9-9.4 kg CO2/kg product according to LULUC scenarios). Correspondingly, 0.5 kg CO2-equiv. per kg of cow milk are emitted as methane, 0.29 kg as N2O and 0.5-0.88 kg as CO2. GHG fluxes from LULUC range between 0.26 kg and 0.64 kg.

There are large ranges in emission intensities found across EU-27 countries. For example, emission intensities from beefrange between 14.2 and 17.4 kg CO2-equiv. per kg product in Austria and the Netherlands, respectively, to values above 40 kg CO2-equiv. per kg product in Cyprus and Latvia. For cow milk, most of the countries that were part of the European Union before 2004 ('old member states') show emission intensities between 1 and 1.4 kg CO2-equiv. per kg product, while those countries accessing the EU after that date ('new member states') show values above 1.5 kg CO2-equiv./kg product (for Scenario II). Net poultry meat emission intensities range between 3.3 kgCO2-equiv. per kg of poultry in Ireland and 17.8 kg in Latvia.

4. Discussion

4.1. Discussion of results

Generally, due to a more efficient digestion process and the absence of enteric fermentation, emissions from pork and poultry meat production are lower than those from ruminant production like beef or sheep and goat meat. Moreover, emissions from poultry meat are lower than those from pork for all gases, which can be explained by a better feed to output relation, different loss factors and in case of energy related emissions lower energy requirements for stables.

Emission intensities differ considerably between countries for all products examined. For beef, the countries with lowest emission intensities are as diverse as Austria and the Netherlands. In the Netherlands, CH4 and N2O emissions are particularly low, which indicates an efficient and industrialized production structure. Austria outbalances higher CH4 emissions by lower fluxes from LUC indicating high self sufficiency in feed production and a high share of grass in the diet. Both countries, however, are characterized by high meat yields. On the other end of the range, meat yields in Latvia are low leading to high emission intensities. Very high fluxes from LUC, as observed e.g. for Latvia and Cyprus, are due to high import shares (in case of Cyprus), or own expansions of agricultural area which is assumed to be at the cost of grasslands (in case of Latvia).

Emission intensities of cow milk seem to be driven by a mixture of effects: High emissions are generally related to low milk yields as in the case of some new member states such as Bulgaria, Romania, Lithuania and Latvia (milk yields around 3500 kg per year). However, high milk yields are often related to the consumption of feed

Fig. 2. Total GHG fluxes of EU-27 livestock products in 2004, calculated with a cradle-to-gate life-cycle analysis with CAPRI (in kg CO2-equiv. per kg of product). (a) beef; (b) sheep and goat meat; (c) pork; (d) poultry meat; (e) cow milk; (f) sheep and goat milk; and (g) eggs.

concentrates. If feed concentrates are imported from overseas, they are frequently accompanied by high emissions from LUC, which can be observed for the Netherlands.

Environmental factors are important determinants for CH4 emissions from manure or emissions from cultivated organic soils. Therefore, we find high emissions in Scandinavian countries with a high share of organic soils or in Mediterranean countries due to high average temperatures, for example for sheep and goat.

Emissions can also be influenced by the implementation of manure management technologies. So, in case of pork and poultry meat, the high N2O emissions in the Netherlands are related to a high implementation rate of NH3-reduction measures increasing N2O emissions substantially (Klimont and Brink, 2004).

For emissions from cattle, digestibility of feed is directly derived from feed intake being an important factor determining CH4 emissions from enteric fermentation. High CH4 emissions generally

indicate a low digestibility of feed, e.g. through high shares of grass or hay in the diet. In some cases the calculation of the digestibility leads to surprising results. That is, even though more energy is required for pasturing (due to the movement of animals), it does not lead to higher methane emissions, because the better digestibility of fresh grass compared to cut grass, silage or hay (NRC, 2001). In contrast, N2O emissions generally increase with the share of solid manure management systems or manure fallen on pastures. Furthermore, we assume that carbon sequestration in grassland is increased in managed grasslands and pastures compared to natural grassland, but decreased in croplands. Since we take natural grassland as the reference situation, in countries with high shares of grass in the diet, the credits offset most of the foregone carbon sequestration for the cultivation of feed crops. In some countries, a net carbon sequestration in grassland has been calculated for beef and sheep and goat products.

The reason for differences is thus a complex interaction between many factors including farming systems, environmental circumstances, and production levels. However, the results suggest that three main factors are important for a country to have low emission intensities: (i) high productivity; (ii) low dependency of imported feed products; and (iii) a high share of pasture in the animal feed diet.

4.2. Discussion of methodology

4.2.1. Land use and land use change

The approach for the calculation of emissions and removals from LUC in some aspects differs to approaches in other studies. First, we do not derive emissions from observed (or projected) LUCs like in FAO (2006), which estimates are based on Wassenaar et al. (2007), or in Zaks et al. (2009), but on land need derived from crop statistics (see Annex). This concept is also applied in Hiederer et al. (2010). The advantage of the selected approach is that agricultural statistics are generally available for all countries, more reliable than land use statistics, and automatically provide crop specific information instead of information on the level of land use aggregates like arable land. Instead of assigning observed LUCs to crops our approach requires to assign observed crop expansion to converted land use sources. Hiederer et al. (2010) solves it with a spatial allocation model, which assigns the observed or expected LUCs to the most suitable places on the land use map. Most studies focus only on soybeans from South America, and are based on simplifying assumptions on the share of imported soybeans from Brazil and Argentina that derives from recently deforested land (e.g. 100% and 10%, respectively in Gavrilova et al., 2010); Hortenhuber et al. (2010) claims that 100% of soybean meal from South America comes from recently cleared land with savannah-type vegetation; Smaling et al. (2008) assigned 50% of deforestation in Brazil to soybean production. In contrast, we decided to focus on the range of possible emission outcomes by defining a minimum and a maximum scenario via respective shares of forest, shrubland or grassland being transformed. Moreover, we do not restrict our analysis to a few important countries and feed products, but we consider all feed exporting countries and most feed products traded on the world market.

A third difference is that we do not assign the emissions of a specific crop, e.g. soybean, to the respective production of the cleared area (like in Hiederer et al., 2010) but to total production of this crop in the respective country. Moreover, we do not use the concept of a depreciation period for which the production on the cleared piece of land is charged with the emissions from clearing (generally most studies use 20 years), but simply assign emissions created for the production of the crop, the period and the country to the total production of the crop in the same period and the same country. The resulting country-specific LUC-factors are more stable,

e.g. an increase from 10 to 20 years observation period would not have a too heavy impact on the results, and complicated dynamic effects can be avoided. In order to avoid unreasonable charges due to specific import patterns we finally create a unique LUC-factor by a weighted average over all producer countries.

One argument against this approach is that emissions should be assigned to those who cause them. However, the assignment to marginal demand would charge the whole burden to countries with an emerging livestock production while leaving countries with a longer importing history free of charge. In contrast, our approach assumes that all buyers of a product share the same responsibility for emissions related to the production independently of historical patterns.

4.2.2. Allocation method

In contrast to many LCA studies our approach does not use the economic allocation to assign emissions to multiple products but an approach based on the physical and causal relationships between products and emissions. For N2O emissions, it has been argued that the N-input is the single most important factor determining the emission strength (Bouwman, 1996; Leip et al., 2005; Skiba and Smith, 2000; Stehfest and Bouwman, 2006) even though it is known that other - mainly environmental and farm management parameters - have significant influence. For CH4 emissions from enteric fermentation, energy requirement is one of the main determining factors. Usually economic allocation leads to about 90% allocation of total emissions from dairy herds to milk, while it is about 85% in the case of physical allocation (Cederberg et al., 2009). However, the number depends significantly on the definition of the dairy herd, which could be the whole herd including fattening animals like in Casey and Holden (2005a) or just the cows with the young calves and the raising of females for cow replacement (see a.o. Castanheira et al., 2010; Haas et al., 2001; Williams et al., 2006). In our case, we allocate between 70% and 98% of dairy cow's emissions to milk if applying the nitrogen allocation, and 72-96% if applying the allocation by energy requirement. Emissions from the total cattle herd (including dairy cows) are allocated at around 50% to milk and 50% to beef.

The use of a physical basis for allocation factors has also been proposed by ISO 14044 (ISO, 2006a), even though later guidelines such as the PAS2050 (BSI, 2008) recommend to apply economic allocation if it is not possible to define sub-processes or to apply system expansion. IDF (2010) argues that it might be difficult to find a common basis for a physical relationship, as soy meal is typically used for its protein content, while soy oil is used for its energy content. This reasoning is valid from a conceptual perspective of a consequential' LCA which tries to quantify marginal effects of products (Schmidt, 2008; Thomassen et al., 2008a; Weidema, 2000) and argues from a viewpoint of "why has this product been produced?" In contrast, the attributional LCA-approach intends to quantify emissions from a retroperspective viewpoint asking "what emissions are caused by this product?" Intuitively, the consequential approach appears to give more accurate information on the impact of a product as the driving force behind its production is the economic value for the producers; most researchers, therefore, implicitly take LCA as a synonymous for 'consequential LCA', and hence the preference for the economic allocation method also for emissions.

Taking the difference between the two approaches serious, however, allocation on a physical basis seems more suitable in an attributional LCA like the present study. First, in order to ensure comparability between emissions, regional differences in prices should not lead to different emissions. In an attributional concept, a farmer who cultivates soybean in a country where biofuels are subsidised should not get higher emissions in the oil product than a farmer who cultivates soybean (under equal conditions) in

a country without extra payment. In a consequential concept, it should. Second, emissions from a product obtained under equal conditions (environment and management) should be assigned with equal emissions and not subjected to price fluctuations in time.

One of the main arguments against allocation on the basis of protein content is the use of multiple products for different purposes, as it is the case for soybean oil (for energy generation, carbon content is important) and soybean cakes or meals (as protein-rich feed). Nguyen et al. (2010) use emissions from avoided palm oil in a substitution approach. In our methodology, emissions per kg of soybean cake and soybean oil are, on average, 1.17 and 0.34 times the value of soybean, respectively.

4.3. Uncertainty

Estimation of GHG emissions is generally subject to large uncertainties, and this is true specifically for estimates of emissions from the agricultural sector (Leip et al., 2005; Leip, 2010; Winiwarter, 2007). The reason is high variability in emission factors, in particular for N2O emissions from agricultural soils, combined with a high cost of measurements, so that reliable emission factors at regional scale are difficult to obtain (Leip et al., 2011c; IPCC, 2006) and a bias in model assumptions difficult to identify (de Vries et al., 2011; Leip et al., 2011d).

In our approach, model assumptions and emission factors are taken from official guidelines and established models (IPCC, 2006; Velthof et al., 2009). IPCC (2006) provides estimates on uncertainties of default emission factors, with uncertainty ranges covering one order of magnitude, as for example for direct N2O emission from agricultural soils. Particular high uncertainty is also associated with the estimation of emissions from carbon sequestration and land use changes (see e.g. Marelli et al., 2011) and we therefore have decided to run three scenarios encompassing the range of expected fluxes.

However, Leip (2010) points out that the uncertainty of activity data is often underestimated. The CAPRI database uses official EU statistics. However, a closer look to those statistics reveals that data are usually collected independently from each other and in many cases they are not consistent. Therefore, CAPRI applies an internal procedure which corrects data automatically if they do not fit together (see Britz and Witzke, 2008; Witzke et al., 2008). In this procedure, constrained estimation techniques are used to fill data gaps or remove inconsistencies or data errors, such as statistical outliers or implausible breaks in time series. As an example, livestock numbers used in CAPRI in some cases (see Leip et al., 2010) diverge from numbers in the official GHG inventories (EEA, 2010).

Similarly, data required for the estimation are not always available or they are not available at the required level of detail. In those cases CAPRI uses estimates from literature or breaks down aggregated numbers to the more detailed level on the basis of available information. For example, default nitrogen excretion data are available from IPCC (2006), but these are not suited for an application in an LCA model, which should reflect regional differences in animal feeding and animal performance. In CAPRI, an animal-budget is used to estimate N-excretion data dynamically (Leip et al., 2011a). An animal budget requires statistical information on grassland yield, which are often missing from statistical sources; errors made in the assumptions used (see Oenema et al., 2007) will propagate into the calculated N-excretion values and related emissions.

4.4. Comparison with other studies

A series of national and regional LCA-studies on specific animal products have been published in the last decade. de Vries and de Boer (2010) and Yan et al. (2011) present a compilation of available

studies and point out that they generally differ in the functional unit, the system boundaries or the allocation approach, and are, therefore, not easily comparable. First, most studies do not consider emissions from LULUC (and if they do, they do it in different ways). Second, some studies use a cradle to farm gate, others a cradle to grave approach (with e.g. raw milk or packaged milk as functional unit, respectively). Thirdly, in contrast to the present study, most studies use an economic or mass allocation method instead of a nutrient or energy related approach. Finally, some studies use national, others regional values and some differentiate between production systems instead of regions.

A detailed assessment of the global dairy sector has been published by the Food and Agriculture Organization (FAO, 2010). Similar to earlier work (FAO, 2006), it is a cradle to grave LCA including emissions from LUC of soybean production in Brazil and Argentina. FAO (2010) estimates emission intensities for different world regions on a per product level. While on global level average emissions are estimated at 2.4 kg CO2-equiv. per kg of milk, for Western and Eastern Europe the value is around 1.4 kg which matches well with the results of the present study. However, while in the present study land use emissions and emissions from farm capital goods were included, and LUC emissions were estimated world wide and for all feed products, FAO (2010) includes post-farm emissions (c. 0.2 kg CO2-equiv. per kg milk) and uses slightly different values for the Global Warming Potential. The contribution of LUC emissions (0.04-0.11 kg CO2-equiv. per kg milk) is significantly lower in FAO (2010). Emissions up to the farm gate (without LULUC), however, seem to be in the same range of around 1.1 kgCO2-equiv. per kg milk in both studies.

Lesschen et al. (2011) present GHG emission profiles for the European livestock sector, calculated with the MITERRA-EUROPE model. As in the present study, a cradle to farm gate LCA has been performed, and basic input data like herd sizes and acreages were obtained from the CAPRI model. The study considers beef, cow milk, pork, poultry meat and eggs, generally uses IPCC (2006) emission factors, but does not consider fluxes from LUC and carbon sequestration. Total emissions amount to 493Tgof CO2-equiv. for the EU, which is about 23% higher than the value presented here (401 Tg of CO2-equiv.), if only emission categories present in both studies and the slightly different GWPs used are considered. On a more detailed scale results partly differ considerably. First, Lesschen et al. (2011) do not include emissions from feed transport, indirect emissions in buildings and machinery, pesticide use and feed processing. Consequently, the emissions from energy use are only one fourth of those estimated in the present study. Second, for emissions from manure management Lesschen et al. (2011) assumes liquid systems without a natural crust cover, while we assume liquid systems with a natural crust cover. As a consequence, in the present study methane emissions from manure management are lower (-48%), while N2O emissions are higher (+309%). For all other categories emissions presented here are equal or lower than in Lesschen et al. (2011). Important differences are: (i) Different allocation methods for manure applied to crops not dedicated to feed production. In the present study, the substitution approach was used and saved emissions for mineral fertilizer application and production are credited to the animal activity. Therefore, emissions from mineral fertilizer production related to animal products are substantially lower in the present study (-64%). (ii) Different N-excretion factors were used. While in the present study these are endogenously calculated with the CAPRI model, and are consistent with feed intake and nitrogen retention values in products (Leip et al., 2011a, 2011b), Lesschen et al. (2011) used fixed national excretion values from the GAINS model (Klimont and Brink, 2004). Furthermore, the MITERRA model did not calculate feed digestibility endogenously based on feed rations but used the IPCC Tier 1 approach also for ruminant emissions from enteric fermentation. (iii) At the product level,

Lesschen et al. (2011) meat is equivalent to 90% of the carcass which increases emissions per kg meat around 10% compared to the present study.

There are also significant differences in the ranking of the countries with respect to estimated emission intensities. For example, Lesschen et al. (2011) estimate that the Netherlands is one of the countries with the highest emissions per kg of beef, while Denmark shows one of the lowest emissions. In the present study the Netherlands have the lowest emissions if LUC and carbon sequestration is not considered, while Denmark is significantly above average. In the case of Sweden and the Netherlands, excretion rates in CAPRI are considerably higher than the GAINS values, which might be explained by high levels of proteins from fodder on arable land in the database.

Most LCA-studies at national or regional level deal with dairy systems in the old member states. In the following we therefore focus on milk. To make the values comparable to our study, we took the numbers without LULUC and tried to harmonize the numbers to a cradle to farm gate approach, applying the functional units of the current study. In case of differentiated production systems we take the whole range. Studies are available for the German region Allgaeu (Haas et al., 2001), Ireland (Casey and Holden, 2005b, 2006), the Netherlands (Thomassen et al., 2009, 2008b), Austria (Hortenhuber et al., 2010), Portugal (Castanheira et al., 2010), Sweden (Cederberg and Mattsson, 2000), and England (Williams et al., 2006). The results are similar to the data calculated with CAPRI (0.83-1.12 kg CO2-equiv./kg milk in old member states, and 1.02-1.61 kg CO2-equiv./kg in the new member states), showing a somewhat wider range of 0.83-1.5 kg CO2-equiv./kg milk.

For Sweden the CAPRI estimates are above the values found in the literature, which is probably due to high feeding values in CAPRI in Denmark and Sweden. For Austria, the results correspond very well: 0.93 kgCO2-equiv. in the present study, compared to a range of 0.83 kg to 1.09kgCO2-equiv. for different locations and production systems in Hortenhuber et al. (2010). For the German region, Portugal and England the differences are between 5 and 10%.

In contrast, for the Netherlands and Ireland differences of 40-50% require a closer look. For the Irish study the main part of the deviation can be explained by the fact that the total herd emissions, including calves, bulls and heifers for fattening, are almost entirely allocated to milk (in fact the study shows results from three different allocation mechanisms, but in none of them more than 15% of total herd emissions would be allocated to meat). In contrast, in the present study only emissions of the dairy activity (time from the first lactation to the slaughtering of the cow) are allocated to milk (just a small share is going to calves), while growing and fattening activities are generally assigned to meat. Most other studies take the cow's emissions from birth to death, and then, usually according to mass or economic value, allocate part of them to milk. This generally leads to results similar to those of the present study. In case of the Dutch study the deviation could be due to the same reason as in the Irish study but the paper unfortunately does not provide all the information in order understand the approach and compare it to the present study.

5. Summary and conclusions

This study presents detailed product-based net emissions of main livestock products (meat, milk and eggs) according to an attri-butional cradle-to-gate life-cycle assessment at regional detail for the whole EU-27 and including emissions from LULUC. For LUC, three scenarios have been designed that should span the range of possible net emissions. Calculations were done with the CAPRI model and the covered gases are CH4, N2O and CO2.

Total GHG fluxes of European Livestock production including LULUC fluxes are estimated between 623 Tg CO2-equiv. and

852TgCO2-equiv., depending on which scenario is used for the calculation of LUC emissions. More than 80% of livestock emissions are created by beef, cow milk and pork production, and more than 80% of agricultural sector's emissions according to IPCC classifications, is due to livestock production. These results are assigned with considerable uncertainty. Particularly data for assessing LUC and changing carbon sequestration are uncertain.

Net emissions per kilogram of carcass of meat are higher for ruminants than for pork and poultry meat. Eggs and milk have a considerably lower carbon footprint per kg of product. The countries with the lowest net product emissions are not necessarily characterized by similar production systems. So, in the case of beef (Scenario II) they are as diverse as Austria and the Netherlands. While the Netherlands save emissions especially with low methane and N2O rates indicating an efficient and industrialized production structure with strict environmental regulations, Austria outbalances the higher methane emissions by lower net emissions from LULUC indicating high self sufficiency in feed production and a high share of grass in the diet. The selection of the LUC scenario, therefore, impacts strongly on the relative performance (in scenario III the Netherlands fall back to average).

The LCA methodology reveals that for the livestock sector the IPCC sector 'agriculture' estimates only around the half of total GHG fluxes caused by EU-27 livestock production up to the farm gate. This suggests that for effective reduction of GHG emissions from livestock production, fluxes occurring outside the agricultural sector need to taken into account. Reduction targets should address both the production side as defined by IPCC sectors and the consumption side. An LCA assessment as presented here could be a basis for such efforts.

Acknowledgements

The authors gratefully acknowledge support from the European Commissions for the GGELS project (contracts AGRI-2008-0245 and AGRI-2009-0296) and the CC-TAME project (FP7 grant no. 212535), underpinning the present study.

Annex. Emissions from land use change

Emissions from land use change are calculated for carbon stock changes in above and below-ground biomass and dead organic matter (BIO + LIT), soil organic carbon stock changes (SOI) and emissions of CH4 and N2O from biomass burning (BUR). We use the Tier 1 methodology proposed by the IPCC (2006). First, the area of land transformed to cropland is estimated based on additional land requirements and transition probabilities. Then, emission intensities LUCFGAS CAT (in the following called LUC-factor) for feed products are calculated according to equation (LUC1)

(LUC1) LUCFcCAS CAT = LUAc x EF£A

LUAc is the expansion of cropland assigned to crop c, in ha per kg; EFCAT is the emission factor for GAS (CO2, CH4, N2O) and CAT (BIO + LIT, SOI, BUR), in kg GAS per ha; LUCF^AS CAT is the emission factor (LUC-Factor) per kg of feed product c for Gas (CO2, CH4, N2O) and CAT (BIO + LIT, SOI, BUR), in kg GAS per kg;

The following sections provide an overview of the methodologies used for the development of emission factors, additional land requirement, and transition probabilities.

A.1. Additional land requirement

A simplified approach was chosen in order to provide an idea of the average yearly requirement of additional land for different crop products. Based on time series of the FAO crop statistics

(http://faostat.fao.org; accession date: 23/03/2010), the change of total cropland area and (the change of) the area for single crops was calculated for a ten year period (1999-2008) in all EU countries and Non-EU country blocks used in the CAPRI model. For those regions where the total cropland area has increased the additional area was assigned to crops by their contribution to area increases. Finally, the area assigned to a certain crop c was divided by the total production of the crop in the region Pc over the same time period, in order to derive the area of cropland expansion per kg of the crop product LUAc.

(LUC2) LUAc =

aic/J2caic x AI

aic is the expansion of the area for crop c (crops with area reduction not considered), in ha; LUAc is the expansion of cropland assigned to crop c, in ha per kg; AI is the total expansion of cropland, in ha; Pc is the total production of crop c, in kg;

zero carbon stock for cropland due to the fact that the biomass is created and removed each year:

(LUC3) EF£°2 = V PLU x cBIO+7l't x shCU :

' LU,CZ — / jt LU,C7

PLU is the probability that new cropland is coming from land use LU in the respective country or country block; CB/f+L17 is the carbon stock of above and below ground biomass and dead organic matter (litter) of land use LU in climate zone CZ in the respective country or country block, in kg C per ha; shC/ is the share of climate zone CZ in area of land use LU in the respective country or country block; EF£jO2+LIT is the CO2-emission factor from above and below ground biomass and dead organic matter (litter) in the respective country or country block per ha of area transformed to cropland, in kg CO2 per ha; pLU is the transition probability corresponding to the respective scenario and the shares of climate zones for different land uses shCZ as described above. 44/12 is conversion factor carbon to CO2.

A.2. Transition probabilities

The transition probabilities from other land uses to cropland pLU are not available and attempts to derive reasonable numbers from satellite data were not successful since global land use map products (i.e. the MODIS or the GlobCover2000 database) are usually not available in form of time series, and the classification error is generally much higher than the actual land use change (see Fritz et al., 2009). Therefore, we decided not to focus on the exact estimation of land use changes but rather define three scenarios which should span the space of possible outcomes. In Scenario I all additional cropland is assumed to come from grassland and savannas, Scenario

II applies a more likely mix of transition probabilities, and Scenario

III can be considered as a maximum emission scenario. The transition probabilities (pLU) for the scenarios II and III are presented in Table A1.

A.3. Emission factors for carbon stock change in above and below ground biomass and dead organic matter

Carbon stock factors are based on IPCC default values and are taken from Carré et al. (2010) for above and below ground biomass (clucz), and from IPCC (2006, see vol. 4. Chapter 2, Table 2.2) for dead organic matter (C^JCZ).

Land use data are based on three sets of land cover data: (1) The Global Land Cover 2000 product (GLC2000) vsl.1 (http://bioval.jrc.ec.europa.eu/products/glc2000/glc2000.php), (2) The GlobCover project (http://ionia1.esrin.esa.int/), and (3) The M3 land cover data from McGill University (Ramankutty et al., 2008). The data set was provided on a 5 min pixel level (Carré et al., 2010). For the calculation of land use change emissions six land use classes were used: Cropland, Grassland, Shrubland, Forest with less than 30% canopy cover, Forest above 30% canopy cover, and Other Land Uses. For each Pixel the distribution of land use classes is known from the above land cover map, complemented by the assignment of each Pixel to one of nine climatic zones (Boreal, Cool Temperate Dry, Cool Temperate Wet, Warm Temperate Dry, Warm Temperate Wet, Tropical Dry, Tropical Moist, Tropical Wet, Tropical Mountain Climate). The exact methodology for the assignment to Climate zones and land use classes is described in Carré et al. (2010). Information on climate and land use on pixel level is then aggregated to the level of those countries and country blocks, which are used in the CAPRI model.

Country specific emissions per hectare of area transformed to cropland are calculated according to equation (LUC3), assuming a

A.4. Emission factors for soil carbon stock change

The default soil carbon stocks on pixels of 5' x 5' are based on the IPCC default values (IPCC, 2006, vol. 4, Chapter 2, Table 2.3) and have been provided by Carré et al. (2010). The soil parameters applied are taken from the Harmonized World Soil Database (HWSD) from IIASA and FAO. For the exact translation of the World Reference Base (WRB) soil types to IPCC classes see Carré et al. (2010). The soil carbon values on pixel level were aggregated to countries, climate zones and land use, using the information described in the preceding section.

The calculation of the soil carbon emissions per hectare of area transformed to cropland is carried out according to the following formulas, based on IPCC (2006, vol. 4, Chapter 2, Eq. (2.25)):

(LUC4) EPS = ]TPlu x SOClu,cz x F^,cz

^2(FUJ,CZ,MC x shLU,MG) MG

5z(F'lu,cz,in x shLU,IN ) - Fcl,c

x shc,MG)

x shc,IN ) x shCCU x

PLU is the probability that new cropland is coming from land use LU in the respective country or country block; SOCLUCZ is the default soil carbon stock of land use LU in climate zone CZ in the respective country or country block, in kg C per ha; F^j CZ is the stock change factor for land use systems of climate zone CZ and land use LU (c = cropland) in the respective country or country block; F¡MJJ CZ MG is the stock change factor for management regime of climate zone CZ, land use LU (c = cropland) and management system MG in the respective country or country block; F'lu cz ¡n is the stock change factor for input of organic matter of climate zone CZ, land use LU (c = cropland) and input category IN in the respective country or country block; shLUMG is the share of management system MG in land use LU in the respective country or country block; shLUjN is the share of input category IN in land use LU in the respective country or country block; shCJ is the share of climate zone CZ in area of land use LU in the respective country or country block; £F^O/L is the CO2-emission factor from the change of soil carbon in the respective

Table A1

Probabilities pLU for new cropland coming from the following land use categories (in percent).

Scenario

Country

Grassland

Shrubland

Forests less than 30% canopy cover

Forests above 30% canopy cover

Europe (EU and Non-EU), USA, Canada, Russia and former Soviet countries, Japan, Australia and New Zealand India, China, Mexico, Morocco, Turkey, other Non-European Mediterranean countries Argentina, Chile, Uruguay, Paraguay, Bolivia, Least developed countries (incl. ACP) Brazil, Venezuela, Rest of South America, all other countries

Europe (EU and Non-EU), USA Canada

Russia and former Soviet countries, Japan, Mexico, Venezuela, Brazil, Chile, Paraguay, Bolivia, Rest of South America, India, Turkey, Least developed countries (incl. ACP)

Australia and New Zealand, Argentina, all other countries

Uruguay

Morocco, other Non-European Mediterranean countries

100 0 0

25 40 50 50

25 10 25 50

0 50 0

0 50 100

50 50 25 0

country or country block per ha of area transformed to cropland, in kg CO2 per ha.

FM, Fl and F¡ are stock factors which increase or decrease the default (equilibrium) carbon stock SOC according to management systems, land use systems and input of organic matter. The values are taken from IPCC (2006, vol. 4, Chapter 5, Table 5.5 and Chapter 6, Table 6.2). shLUMG, shLUjN are country specific shares of management systems and input categories by land uses. Due to a lack of data on management and input they are based on a few simple regional assumptions guaranteeing that carbon stocks do not deviate strongly from default values.

A.5. Emission factors (CH4 and N2O) for biomass burning

Our calculation follows a Tier 1 approach of the IPCC guidelines (IPCC, 2006, vol. 4, Chapter 2) and using generally default values. The general formula is:

(LUC5) EFGUARS shB!UR x Plu x FUELlu¡cz x CFluxz x EF^cz

x shCU

shfUR is the share of the cleared area in land use LU which is burned in the respective country or country block; PLU is the probability that new cropland is coming from land use LU in the respective country or country block; FUELlUcz is the dead organic matter and live biomass by land use LU and climate zone CZ, in tonnes of dry matter per ha; CFlUcz is the combustion factor by land use LU and climate zone CZ, in tonnes of dry matter per ha; EF¡UAsCz are emission factors from Burning for GAS (CH4, N2O) by land use LU and climate zone CZ, in kg gas per kg dry matter burnt; shCU is the share of climate zone CZ in area of land use LU in the respective country or country block; EF^fiS are emission factors from Burning for GAS (CH4, N2O) in the respective country or country block per ha of area transformed to cropland, in kg gas per ha.

For the share of area burnt shBUS a value of 50% is assumed for forest and shrubland, and a value of 35% for grassland converted to cropland. This corresponds to the default values recommended by the IPCC guidelines (IPCC, 2006, vol. 4, Chapter 5, p. 5.29). Similarly, the values for dead organic matter and live biomass values (FUELlUcz), indicating the amount of fuel that can be burnt, the applied combustion factors (CFLUCZ), which measure the proportion of the fuel that is actually combusted and varies with the size and composition of the fuel, the moisture content and the type of

fire, and the default emission factors EFCH^ and EFjNjOz are taken from IPCC (2006, vol. 4, Chapter 2, Tables 2.4-2.6). In case ofbiomass the values for the land use category "Forest less than 30% canopy cover" are generally 20% of the default values for the respective forest category.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.agee.2011.12.015.

References

Bouwman, A., 1996. Direct emission of nitrous oxide from agricultural soil. Nutr.

Cycl. Agroecosyst. 46,53-70. Britz, W., Verburg, P.H., Leip, A., 2011. Modelling of land cover and agricultural change in Europe: combining the CLUE and CAPRI-Spat approaches. Agric. Ecosyst. Environ. 142, 40-50, doi:10.1016/j.agee.2010.03.008. Britz, W., Witzke, H.-P. (Eds.), 2008. CAPRI Model Documentation 2008: Version 2 p. Institute for Food and Resource Economics, University of Bonn, Bonn, available at: http://www.capri-model.org/docs/caprLdocumentation.pdf. BSI, 2008. PAS 2050:2008. Specification for the Assessment of the Life Cycle Greenhouse Gas Emissions of Goods and Services. British Standards Institution, London.

Carré, F., Hiederer, R., Blujdea, V., Koeble, R., 2010. Background Guide forthe Calculation of Land Carbon Stocks in the Biofuels Sustainability Scheme. EUR 24573 EN. European Union, Luxembourg. Casey, J.W., Holden, N.M., 2005a. Analysis of greenhouse gas emissions from the average Irish milk production system. Agric. Syst. 86 (1), 97-114, doi:10.1016/j.agsy.2004.09.006. Casey, J.W., Holden, N.M., 2005b. The relationship between greenhouse gas emissions and the intensity of milk production in Ireland. J. Environ. Qual. 34 (2), 429.

Casey, J.W., Holden, N.M., 2006. Quantification of GHG emissions from sucker-beef

production in Ireland. Agric. Syst. 90 (1-3), 79-98. Castanheira, É.G., Dias, A.C., Arroja, L., Amaro, R., 2010. The environmental performance of milk production on a typical Portuguese dairy farm. Agric. Syst. 103 (7), 498-507, doi:10.1016/j.agsy.2010.05.004. Cederberg, C., Mattsson, B., 2000. Life cycle assessment of milk production - a comparison of conventional and organic farming. J. Cleaner Prod. 8 (1), 49-60, doi:10.1016/S0959-6526(99)00311-X. Cederberg, C., Sonesson, U., Henriksson, M., Sund, V., Devis, J., 2009. Greenhouse Gas Emissins from Swedish Production of Meat, Milk and Eggs 1990 and 2005. SIK Report Nr. 793 2009. The Swedish Institute for Food and Biotechnology. de Vries, M., de Boer, I.J.M., 2010. Comparing environmental impacts for livestock products: A review of life cycle assessments. Livestock Sci. 128 (1-3), 1-11, doi:10.1016/j.livsci.2009.11.007. de Vries, W., Leip, A., Reinds, G.J., Kros, J., Lesschen, J.P., Bouwman, a.F., 2011. Comparison of land nitrogen budgets for European agriculture by various modeling approaches. Environ. Pollut. 159,3254-3268, doi:10.1016/j.envpol.2011.03.038. EEA, 2010. Annual European Community Greenhouse Gas Inventory 1990-2008 and Inventory Report 2010. Submission to the UNFCCC Secretariat. European Environment Agency, Available at: http://www.eea.europa.eu/data-and-maps/data/ds_resolveuid/69c4be530d66541bbadc5dae4cd7b33c.

FAO, 2006. livestock's Long Shadow - Environmental Issues and Options. Food and Agriculture Organization of the United Nations, Rome.

FAO, 2010. Greenhouse Gas Emissions from the Dairy Sector. A Life Cycle Assessment. Food and Agriculture Organization of the United Nations.

Fritz, S., McCallum, I., Schill, C., Perger, C., Grillmayer, R., Achard, F., Kraxner, F., Obersteiner, M., 2009. Geo-Wiki.Org: the use ofcrowdsourcing to improve global land cover. Remote Sens. 1,345-354, doi:10.3390/rs1030345.

Galloway, J., Dentener, F., Burke, M., Dumont, E., Bouwman, A.F., Kohn, R.A., Mooney, H.A., Seitzinger, S., Kroeze, C., 2010. The impact of animal production systems on the nitrogen cycle. In: Steinfeld, H., Mooney, H., Schneider, F., Neville, L. (Eds.), Livestock in a Changing Landscape. Volume 1. Drivers, Consequences and Responses. Island Press, Washington, USA, pp. 83-95.

Garnett, T., 2009. Livestock-related greenhouse gas emissions: impacts and options for policy makers. Environ. Sci. Policy 12 (4), 491-503, doi:10.1016/j.envsci.2009.01.006.

Gavrilova, O., Jonas, M., Erb, K., Haberl, H., 2010. International trade and Austria's livestock system: direct and hidden carbon emission flows associated with production and consumption of products. Ecol. Econ. 69 (4), 920-929, doi:10.1016/j.ecolecon.2009.11.015.

Haas, G., Wetterich, F., Köpke, U., 2001. Comparing intensive, extensified and organic grassland farming in southern Germany by process life cycle assessment. Agric. Ecosyst. Environ. 83 (1-2), 43-53, doi:10.1016/S0167-8809(00)00160-2.

Hiederer, R., Ramos, F., Capitani, C., Koeble, R., Blujdea, V., Gomez, O., Mulligan, D., Marelli, L., 2010. Biofuels: A New Methodology to Estimate GHG Emissions from Global Land Use Change. EUR 24483 EN. European Union, Luxembourg.

Hörtenhuber, S., Lindenthal, T., Amon, B., Markut, T., Kirner, L., Zollitsch, W., 2010. Greenhouse gas emissions from selected Austrian dairy production systems -Model calculations considering the effects of land use change. Renew. Agric. Food Syst. 25 (4), 316-329, doi:10.1017/S1742170510000025.

IDF, 2009. Environmental/ecological impact of the dairy sector: literature review on dairy products for an inventory of key issues. List of environmental initiatives and influences on the dairy sector. Bull. Int. Dairy Fed. 436.

IDF, 2010. A common carbon footprint approach for dairy. The IDF guide to standard lifecycle assessment methodology for the dairy sector. Bull. Int. Dairy Fed. 445.

IPCC, 1997. Revised 1996 IPCC Guidelines for National Greenhouse Gas Inventories. IPCC/OECD/IEA, Paris, France.

IPCC, 2003. In: Penman, J., Gyatarsky, M., Hiraishi, T., Krug, T., Kruger, D., Pipatti, R., Buendia, L., Miwa, K., Ngara, T., Tanabe, K., Wagner, F. (Eds.), Good Practice Guidance for Land Use, Land-Use Change and Forestry. IPCC National Greenhouse Gas Inventories Programme, Hayama, Japan.

IPCC, 2006. Prepared by the National Greenhouse Gas Inventories Programme, Eggleston, H.S., Buendia, L., Miwa, K., Ngara, T., Tanabe, K. (Eds.), 2006 IPCC Guidelines for National Greenhouse Gas Inventories. IGES, Japan.

ISO, 2006a. ISO 14044:2006(E). Environmental Management - Life Cycle Assessment

- Requirements and Guidelines. International Standard Organization, Geneva, Switzerland.

ISO, 2006b. ISO 14040:2006(E). Environmental Management - Life Cycle Assessment - Principles and Framework. International Standard Organization, Geneva, Switzerland.

Jansson, T., Heckelei, T., 2011. Estimating a primal model of regional crop supply in the European Union. J. Agric. Econ. 62, 137-152, doi:10.1111/j.1477-9552.2010.00270.x.

Klimont, Z., Brink, C., 2004. Modelling of Emissions of Air Pollutants and Greenhouse Gases from Agricultural Sources in Europe, IIASA Interim Report 1R-04-048. Interim Report 1R-04-048. International Institute for Applied Systems Analysis, Laxenburg, Austria.

Kränzlein, T., 2009. Economic Monitoring of Fossil Energy Use in EU Agriculture. ARTSchriftenreihe 11, December 2009. Forschungsanstalt Agroscope Reckenholz-Tänikon ART, Tänikon, Ettenhausen, Switzerland.

Leip, A., 2010. Quantitative quality assessment of the greenhouse gas inventory for agriculture in Europe. Clim. Change 103 (1-2), 245-261, doi:10.1007/s10584-010-9915-5.

Leip, A., Dämmgen, U., Kuikman, P., van Amstel, A.R., 2005. The quality of European (EU-15) greenhouse gas inventories from agriculture. Environ. Sci. 2(2-3), 177-192, doi:10.1080/15693430500393177.

Leip, A., Weiss, F., Wassenaar, T., Perez, 1., Fellmann, T., Loudjani, P., Tubiello, F., Grandgirard, D., Monni, S., Biala, K., 2010. Evaluation of the Livestock Sector's Contribution to the EU Greenhouse Gas Emissions (GGELS)

- Final Report. European Commission, Joint Research Centre, Available at: http://afoludata.jrc.ec.europa.eu/index.php/dataset/detail/236.

Leip, A., Weiss, F., Britz, W., 2011a. Agri-environmental nitrogen indicators for EU27. 1n: Filchman, G. (Ed.), Bio-Economic Models Applied to Agricultural Systems. An 1ntegrated Approach to Relations Between Agriculture, Environment and Natural Resources - Tools for Policy Analysis. Springer, pp. 109123.

Leip, A., Britz, W., De Vries, W., Weiss, F., 2011b. Farm, land, and soil nitrogen budgets for agriculture in Europe calculated with CAPRI. Environ. Pollut., doi:10.1016/j.envpol.2011.01.040.

Leip, A., Busto, M., Winiwarter, W., 2011c. Developing spatially stratified N2O emission factors for Europe. Environ. Pollut. 159, 3223-3232, doi:10.1016/j.envpol.2010.11.024.

Leip, A., Busto, M., Corazza, M., Bergamaschi, P., Koeble, R., Dechow, R., Monni, S., de Vries, W., 2011d. Estimation of N2O fluxes at the regional scale: data, models, challenges. Curr. Opin. Environ. Sustain. 3, 328-338, doi:10.1016/j.cosust.2011.07.002.

Lesschen, J.P., van den Berg, M., Westhoek, H.J., Witzke, H.-P., Oenema, O., 2011. Greenhouse gas emission profiles of European livestock sectors. Anim. Feed Sci. Technol. 166-167,16-28, doi:10.1016/j.anifeedsci.2011.04.058.

Marelli, L., Mulligan, D., Edwards, R., 2011. Critical Issues in Estimating ILUC Emissions, EUR 24816. Joint Research Center, European commission, 64 pp, available at: http://re.jrc.ec.europa.eu/bf-tp/download/EU_report_24816.pdf.

Nguyen, T.L.T., Hermansen, J.E., Mogensen, L., 2010. Environmental consequences of different beef production systems in the EU. J. Cleaner Prod. 18 (8), 756-766, doi:10.1016/j.jclepro.2009.12.023.

NRC, 2001. Nutrient Requirements of Dairy Cattle. National Academy Press, Washington, DC.

Oenema, O., Oudendaag, D., Witzke, H.-P., Monteny, G.-J., Velthof, G., Pietrzak, S., Pinto, M., Britz, W., Schwaiger, E., Erisman, J.W., de Vries, W., van Grinsven, J.J.M., Sutton, M., 2007. Integrated Measures in Agriculture to Reduce Ammonia Emissions. Final Summary Report. Service contract "Integrated Measures in Agriculture to Reduce Ammonia Emissions". Alterra-Report 1663.4. Alterra, Wageningen, The Netherlands, available at: http://www.scammonia.wur.nl/NR/rdonlyres/E46888D3-6C8E-4BCC-90F1-154D2B270654/44493/finalreportASC25062007.pdf.

Perez-Dominguez, I., Britz, W., Holm-Müller, K., 2009. Trading schemes for greenhouse gas emissions from European agriculture: a comparative analysis based on different implementation options. Rev. Agric. Environ. Stud. 90, 287-308.

Ramankutty, N., Evan, A.T., Monfreda, C., Foley, J.A., 2008. Farming the planet: 1. Geographic distribution of global agricultural lands in the year 2000. Glob Bio-geochem. Cycles 22,1-19.

Schmidt, J.H., 2008. System delimitation in agricultural consequential LCA. Outline of methodology and illustrative case study ofwheat in Denmark. Int. J. Life Cycle Assess. 13 (4), 350-364, doi:10.1007/s11367-008-0016-x.

Skiba, U., Smith, K.A., 2000. The control of nitrous oxide emissions from agricultural and natural soils. Chemosphere - Glob. Change Sci. 2 (3-4), 379-386, doi:10.1016/S1465-9972(00)00016-7.

Smaling, E.M.A., Roscoe, R., Lesschen, J.P., Bouwman, A.F., Comunello, E., 2008. From forest to waste: assessment of the Brazilian soybean chain, using nitrogen as a marker. Agric. Ecosyst. Environ. 128 (3), 185-197, doi:10.1016/j.agee.2008.06.005.

Stehfest, E., Bouwman, A.F., 2006. N2 O and NO emissions from agricultural fields and soils under natural vegetation: summarizing available measurement data and modelling of global annual emissions. Nutr. Cycl. Agroecosyst. V74 (3), 207-228, doi:10.1007/s10705-006-9000-7.

Steinfeld, H., Wassenaar, T., 2007. The role of livestock production in carbon and nitrogen cycles. Ann. Rev. Environ. Resour. 32, 15.1-15.24, doi:10.1146/annurev.energy.32.041806.143508.

Steinfeld, H., Mooney, H.A., Schneider, F., 2010. Livestock in a Changing Landscape, Volume 1: Drivers, Consequences and Responses. Island Press.

Sutton, M., Oenema, O., Erisman, J.W., Leip, A., van Grinsven, H., Winiwarter, W., 2011. Too much of a good thing. Nature 472,159-161, doi:10.1038/472159a.

Thomassen, M.A., Dalgaard, R., Heijungs, R., de Boer, I., 2008a. Attributional and consequential LCA of milk production. Int. J. Life Cycle Assess. 13 (4), 339-349, doi:10.1007/s11367-008-0007-y.

Thomassen, M.A., Dolman, M.A., Van Calker, KJ., De Boer, I.J.M., 2009. Relating life cycle assessment indicators to gross value added for Dutch dairy farms. Ecol. Econ. 68 (8-9), 2278-2284, doi:10.1016/j.ecolecon.2009.02.011.

Thomassen, M.A., van Calker, K.J., Smits, M.C.J., lepema, G.L., de Boer, I.J.M., 2008b. Life cycle assessment of conventional and organic milk production in the Netherlands. Agric. Syst. 96 (1-3), 95-107, doi:10.1016/j.agsy.2007.06.001.

Tukker, A., Goldbohm, R.A., de Koning, A., Verheijden, M., Kleijn, R., Wolf, O., Pérez-Domínguez, l., Rueda-Cantuche, J.M., 2011. Environmental impacts of changes to healthier diets in Europe. Ecol. Econ. 70, 1776-1788, doi:10.1016/j.ecolecon.2011.05.001.

Velthof, G., Oudendaag, D., Witzke, H.-P., Asman, W.A.H., Klimont, Z., Oenema, O., 2009. Integrated assessment of nitrogen emissions from agriculture in EU-27 using MITERRA-EUROPE. J. Environ. Qual. 38, 402-417, doi:10.2134/jeq2008.0108.

Wassenaar, T., Gerber, P., Verburg, P.H., Rosales, M., Ibrahim, M., Steinfeld, H., 2007. Projecting land use changes in the neotropics: the geography of pasture expansion into forest. Glob. Environ. Change 17, 86-104, doi:10.1016/j.gloenvcha.2006.03.007.

Weidema, B., 2000. Avoiding Co product allocation in life cycle assessment. J. lnd. Ecol. 4(3), 11-33, doi:10.1162/108819800300106366.

Williams, A.G., Audsley, E., Sanders, D.L., 2006. Determining the Environmental Burdens and Resource Use in the Production of Agricultural and Horticultural Commodities. Main Report of the Defra Research Project 1SO205. Cranfield University and Defra.

Winiwarter, W., 2007. National greenhouse gas inventories: understanding uncertainties versus potential for improving reliability. Water Air Soil Pollut.: Focus 7 (4), 443-450, doi:10.1007/s11267-006-9117-3.

Witzke, H.-P., Tonini, A., Zintl, A., 2008. The Common Agricultural Policy SIMulation (CAPS1M) Model: Database for Agricultural Sector Modelling, EUR 23381 EN -2008. European Commission, Luxembourg.

Yan, M.-J., Humphreys, J., Holden, N.M., 2011. An evaluation of life cycle assessment of European milk production. J. Environ. Manage. 92 (3), 372-379, doi:10.1016/j.jenvman.2010.10.025.

Zaks, D.P.M., Barford, C.C., Ramankutty, N., Foley, J.A., 2009. Producer and consumer responsibility forgreenhousegas emissions from agricultural production: a perspective from the Brazilian Amazon. Environ. Res. Lett. 4 (4), 044010, doi:10.1088/1748-9326/4/4/044010.