Impacts of European livestock production: nitrogen, sulphur, phosphorus and greenhouse gas emissions, land-use, water eutrophication and biodiversity

We estimated emissions of CO₂, CH₄, N₂O, NH₃, NO_x, SO₂, and N and P losses to the hydrosphere related to livestock production in Europe and interpreted the results in the light of the five major environmental threats. This was done by (i) extending the LCA approach in the agro-economic Common Agricultural Policy Regionalised Impact (CAPRI) modelling system (Britz and Witzke 2012)¹¹ , developed for the assessment of GHG emissions and N footprints, to provide cradle-to-farm gate LCA data for reactive nitrogen, in combination with (ii) model results for other compounds (P and SO₂) to provide a comprehensive picture of the environmental impact of EU livestock production. Table 1 gives an overview of emission sources quantified indicating how they are linked to livestock production systems. As the table indicates, emissions caused by livestock rearing were calculated for EU27 only. Emissions from imported livestock products were not included in the analysis, while further emissions associated with cultivation, processing and transport of (non-forage) feed might occur within or outside the EU.

The emissions of GHGs (N₂O and CH₄) and reactive nitrogen (N_r) to air (NH₃, NO_x) and water (N leaching and runoff) and other N flows (such as trade of food and feed) were estimated with CAPRI, using the cradle-to-gate LCA approach implemented for the estimation of GHG emissions (Weiss and Leip 2012) and further extended to include the estimation of N footprints per product group (Leip et al 2014b). Here we present new data that have been calculated when further extending the model with flows of all N_r. All new data presented here include emissions occurring within Europe and outside the EU territory for imported feed and/or land use change (LUC; LUC emissions were calculated using the land use transition probabilities of scenario II in Weiss and Leip (2012), table A1). Also included are credits for carbon sequestration in managed grassland as well as emissions due to foregone carbon calculation in croplands according to Weiss and Leip (2012). To analyze international N flows to and from EU27, the detailed trade data matrix of the trade module in the FAOSTAT database (FAOSTAT 2014) was used taking into account the N content to the 504 commodities involved (vegetal, animal and fiber products) (Lassaletta et al 2014). Internal flows among EU 27 countries were calculated and subtracted. The net N import to EU27 for each world country was estimated by the difference between total imports and exports.

A general description of the CAPRI model and its relevant modules can be found in Britz et al (2010), Jansson and Heckelei (2011), Leip et al (2011b, 2011d), Britz and Wizke (2012) and Perez-Dominguez et al (2012).

In the CAPRI LCA module (Weiss and Leip 2012, Leip et al 2014b, Westhoek et al 2015), total agricultural emissions E_agri were estimated as the sum of flows caused by agricultural production activities, plus emissions caused in earlier phases of the products life cycle, including energy use or land use change. Supplementary Information S1 gives detailed results of the N-LCA for the main six vegetable and six livestock product groups to which the data were aggregated. Differently from Weiss and Leip (2012) and in accordance to Leip et al (2014b) the allocation of flows from primary crop products to secondary products (e.g. soya to soybean oil and soybean cakes) is done by mass. The allocation of emissions from feed production to specific livestock products makes use of the animal budget module in CAPRI where energy and protein requirements are matched with domestic and imported feed supply, and data on farm expenditures for feed (Britz and Witzke 2012, Leip et al 2011d). In a first step, emissions from crop activities are converted into emission intensities and allocated to animal activities and in a second step to animal products (Weiss and Leip 2012).

For our purpose, E_agri was divided into emissions related to livestock production E_lvst and those related to production of crops for other purposes (food, fuel, fibre) E_crop. E_lvst includes emissions from livestock production systems E_anim (e.g. CH₄ emissions from enteric fermentation, emissions from energy use for milking etc) plus the emissions from feed production, transport and processing E_feed.

For each crop product, allocation to food or feed is done on the basis of the market balance which is available in CAPRI at the national level (see supplementary information S2). The share of total quantities of flows allocated to food crops was calculated on the basis of total domestic production (gross production) minus the quantity used for feed, while the share of the flow used to feed is allocated to livestock products based on feed intake quantities. Emission sources considered are given in table 2.1 by Westhoek et al (2015).While this study is restricted to quantifying emissions from EU27 agriculture, E_feed includes both emissions from domestic feed production E_feedeu and emissions from imported feed products E_feedrow:

$$\begin{eqnarray*}{E}_{{\rm{agri}}} & = & {E}_{{\rm{lvst}}}+{E}_{{\rm{crop}}}\\ & = & {E}_{{\rm{feedrow}}}+{E}_{{\rm{feedeu}}}+{E}_{{\rm{anim}}}+{E}_{{\rm{crop}}}.\end{eqnarray*}$$

The contribution to global warming was assessed as the sum of direct GHG emissions (CO₂, N₂O and CH₄), indirect N₂O emissions and C sequestration. Direct N₂O and CH₄ emissions were from CAPRI-LCA, using a global warming potential of 298 kg CO₂eq (kg N₂O)⁻¹ and 25 kg CO₂eq (kg CH₄)⁻¹ (IPCC 2007). Indirect N₂O emissions were estimated as 1% of the emitted N (IPCC 2006). Indirect C sequestration in forests was calculated using an average carbon uptake (sequestration) factor of 35 kg C per kg N deposited for European boreal and temperate forests (de Vries et al 2014, 2009) and a fraction of 0.25 kg N deposited on forest per kg N emitted, based on EU27 total NH₃–N emissions from agriculture and NH₃–N deposition on forests (de Vries et al 2011b). Other 'cooling effects'(Butterbach-Bahl et al 2011) were not included in the quantification.

The contribution to air quality impacts was assessed on the basis of the sum of NH₃ and NO_x emissions, as calculated with CAPRI N-LCA.

Contribution to soil acidification was assessed on the basis of the sum of SO₂, NH₃ and NO_x emissions. NH₃ and NO_x emissions were calculated with CAPRI N-LCA. Total SO₂ emissions were obtained from the EDGAR data base (European Commission 2011). SO₂ emissions caused by agricultural activities were approximated by the ratio of CO₂ emissions from agriculture related to energy use (Weiss and Leip 2012) and total energy CO₂ emissions in EU27 (EEA 2011). This gives a share of about 6%, which is associated with livestock products (4%) and vegetable products (2%), in accordance with the global estimate of FAO (FAO 2006).

The contribution of agricultural sources to emissions of acidifying substances was estimated on the basis of acidity equivalents (Schöpp and Posch 2003). This method converts emissions of S, NO_x and NH_y to acidity equivalents on the basis of the molecular weight m (64, 46, and 17 for SO₂, NO_x, and ${{{\rm{NH}}}_{4}}^{+}$ , respectively) and the charge per mole z (−2, −1, and +1 for ${{{\rm{SO}}}_{4}}^{2-}$,${{{\rm{NO}}}_{3}}^{-},$ and ${{{\rm{NH}}}_{4}}^{+},$ respectively) to get the conversion factors 0.03125 Geq (Gg SO₂)⁻¹, 0.02174 Geq (Gg NO₂)⁻¹, and 0.05882 Geq (Gg NH₃)⁻¹.

The contribution of agriculture, livestock and feed to loss of relative mean species abundance (MSA) was estimated using shares of land use and emissions of NH₃, NO_x, and net GHG exchange, accounting for C sequestration, as calculated with the CAPRI-LCA. The data are linked to the estimates of the effect of the main drivers for biodiversity loss as calculated with the GLOBIO-model (Alkemade et al 2009, Kram and Stehfest 2012, van Vuuren et al 2015). This model gives an absolute loss of 65% MSA caused by land conversion into arable, grazing and forestry (35%, 15% and 14%, respectively), and to pressures such as N deposition (2%), climate change (3%) and land fragmentation (30%) (for details see supplementary information S3).

The contribution of agriculture, livestock and feed to N_r losses to the hydrosphere was derived from the results of CAPRI N-LCA. Contribution of livestock and feed to dissolved inorganic phosphorus (DIP) losses has been calculated by applying the share of P in fertilizers (mineral fertilizer and manure) per crop from CAPRI LCA to Global NEWS results on total and agricultural flows of DIP (Mayorga et al 2010). We were unable to quantify the role of agriculture in the load of particulate phosphorus (PP).

A quantification of the impact of N and P losses was done by combining an analysis of potential risk of eutrophication, based on the Indicator for Coastal Eutrophication Potential (ICEP, Garnier et al 2010, Billen et al 2011) with the estimation of livestock contribution to river nutrient loads provided by the model GREEN (Grizzetti et al 2012) in the different European coastal areas (see supplementary information S4).

N flows from EU27 to other world regions and vice versa for the year 2004 are illustrated in figure 2. Much of the proteins grown in Europe are used to feed livestock. From a total of 16.4 Tg N produced on agricultural land in the year 2004 only 2.4 Tg N yr⁻¹ (about 15%) were supplied for direct human consumption or further processing. Most of it was used as animal feed (8.8 Tg N yr⁻¹ or 54%) or returned to the soil as crop residue (5.1 Tg N yr⁻¹ or 31%). Furthermore, we estimate that livestock received 4.2 Tg N yr⁻¹ from imports or industry (Leip et al 2014b, see also details in supplementary information S2).

Zoom In Zoom Out Reset image size

According to FAO trade statistics, EU27 was in 2004 a net importer of agricultural products with soybean products for animal feed produced in Argentina, Brazil and USA representing 84% of the total net imports of EU27 (see figure 2; for a comparison of EU27 and global agricultural structure and emissions see supplementary information S5) that entails significant trade of embodied cropland surface (MacDonald et al 2015). According to calculations based on FAOSTAT (2014) data, in 2004 about 70% of the European livestock production was used for intra-national consumption and 18%–27% (respectively for chicken and cattle meat, expressed in N) was traded between EU27 countries with significant associated embodied GHG emissions (Caro et al 2014). The EU was thus close to self-sufficiency for meat and dairy products, but the share of pig meat production was much higher than in the rest of the world, while the share of ruminant meat was significantly lower (22% versus 29% globally).

Table 2 shows the results for total agricultural emissions from the EU27 agricultural sector and emissions related to livestock production, feed production and imported feed. Values are provided for NH₃, NO_x, SO₂, the combined effect of the three pollutants converted to acidity equivalents, GHG emissions (CH₄, N₂O, and CO₂), C-sequestration, water pollution by emissions of N and P (as dissolved inorganic P, DIP), the land use and the contribution to the loss of the MSA. Values reported in table 2 refer to the year 2004. Detailed results of the LCA calculation (N_r and GHG emissions) are given in supplementary information S1.

Total agricultural emissions as compared with the total EU27 emissions from all sources (Leip et al 2011a) are given in table 3. Both total and agricultural emissions refer to emissions from EU27 territory. Therefore, agricultural emissions in table 3 do not include emissions associated with imported feed (see table 2).

Air quality

Global warming

Soil quality

Terrestrial biodiversity

Quality of inland and coastal water

For N, combining NH₃ + NO_x emissions, our estimate of 2.6 Tg N yr⁻¹ from agricultural sources using the CAPRI model (excluding energy related NO_x) is 19% and 7% smaller than official estimates of the European Union of 3.2 Tg N yr⁻¹ (EEA 2014) and the estimate of the MITERRA model of 2.8 Tg N yr⁻¹ (Westhoek et al 2014), respectively. The estimated total N excretion in CAPRI, at 8.9 Tg N yr⁻¹, is 88% of the official estimate in EEA (2014, EEA 2014). This difference can be explained by the fact that CAPRI calculates N excretion on the basis of a consistent IPCC Tier 2 approach (animal budget, Leip et al 2011b, IPCC 2006) across all countries, while national inventories in Europe are constructed with a large variety of methods and data quality (Leip 2010). National estimates would be of higher quality than CAPRI estimates from countries with good data (Leip et al 2014b), but some countries still need to improve their methodology (EEA 2014). Furthermore, CAPRI uses ammonia abatement measures from the GAINS model (Klimont and Winiwarter 2011) which may not have been considered in national inventories (Leip et al 2010).

In comparison with 4.4 Tg N yr⁻¹ agricultural NH₃ emissions in the EDGAR data base, our estimate of 2.6 Tg N yr⁻¹ for EU27, excluding emissions from imported feed, is lower. The reason for this might be the lower excretion estimates, although the NH₃ emissions are in line with estimates by the MITERRA model.

While 85% of NO_x emissions were related to energy use, only 0.07 Tg N yr⁻¹ were from non-energy sources. A quality check of the total agricultural emission estimate for NO_x is difficult as no comparable study exists including both NO_x budget flows and NO_x emissions related to energy consumption in agricultural systems. Energy consumption in agriculture is calculated in CAPRI with a dedicated energy module which is also used for GHG emission estimates (Kempen and Kraenzlein 2008, Weiss and Leip 2012); this is also the basis of the estimated contribution of SO₂ emissions. The share of agricultural NO_x and NH₃ to total emissions is within −6% to +16% of earlier estimates (Leip et al 2011a).

Our estimate for the share of agricultural GHG emissions to total GHG emissions based on table 3 is about 13%. It ranges between the value of the official EU GHG inventory (10%, EEA 2014) and other estimates on the shares of agriculture or even livestock production on total GHG emissions (Gerber et al 2013, FAO 2006, Weiss and Leip 2012). The official GHG inventory considers only emissions reported in the agriculture sector, whereas LCA studies also include emissions from Land Use Change (LUC) and from imported feeds, which amounted to 39% of total agricultural emissions (see table 2).

Our estimate for N₂O emissions from agricultural soils is considerably lower than official estimates; a comparison of N₂O emissions between various models (de Vries et al 2011b) showed overall satisfying agreement. No methodology is able to capture the huge variability of N₂O emissions caused by changing soil and climate conditions. In view of the general lack of experimental observations, even process-based models are not able to achieve a closer match than independent calculations using inverse methods (Leip et al 2011c).

LUC is certainly one of the most difficult sources to quantify, as it requires data (or good assumptions) on how much LUC is occurring as a consequence of EU agricultural and livestock production, as well as what kind of LUC is triggered. Indeed, the debate on the best method to estimate LUC emissions from agricultural products is still ongoing (European Commission 2013). The method developed in the CAPRI model (Leip et al 2010, Weiss and Leip 2012) was based on the assumptions that the agricultural market is very fluid and no differentiation between direct and indirect LUC is possible. The approach considers only LUC linked to an expansion of harvested area, very similar to the methods proposed by recent guidelines (Food SCP RT 2013, leap 2014). We used unique LUC factors for imports from a country outside the EU as weighted average for all importing countries accounting for globally connected and substitutable trade flows.

We are aware of the debate on the permanency of carbon sequestration in grassland (Smith 2014), however the approach by Weiss and Leip (2012) is based on the observation that enhanced carbon sequestration rates in grassland are observed also after the 20 years equilibrium time usually used by IPCC methodologies (IPCC 2006), which is also consistent with recent simulations with the CENTURY model (Lugato et al 2014).

Carbon sequestration in forests has been estimated earlier at the scale of the EU27 for the year 2000 by multiplying an estimated N deposition caused by agricultural NH₃ emissions of 0.61 Tg N yr⁻¹ with a C response of 50 kg C per kg N deposited leading to a C sequestration near 30 Tg C yr⁻¹ or 112 Tg CO₂ yr⁻¹ (de Vries et al 2011a), being very close to our estimate of 104 Tg CO₂ yr⁻¹. In our study the estimated N deposition was larger (0.82 Tg N yr⁻¹) while the C:N response was estimated at 35 kg C per kg N deposited.

Estimates of agricultural N-leaching range from 2.0 to 5.7 Tg N yr⁻¹ (de Vries et al 2011b), the higher value being also found in the EU GHG inventory (EEA 2014). Possible reasons for these differences—in addition to those already discussed—are available calibration data for nitrate concentrations which might neglect flows of organic nitrogen to water, and the split of total N between the highly uncertain N₂ emissions and N leaching/runoff (de Vries et al 2011b). Our estimate of nitrate leaching at 5.4 Tg N yr⁻¹ is consistent with the estimate of the European Nitrogen Budget (Leip et al 2011a) which is used in table 3 for total N input to water. Although livestock dominated overall agriculture P flows (73%), agriculture is responsible for only 10% of the total riverine P export (table 3). This is because point sources from human wastewater dominate (accounting for 0.21 Tg P yr⁻¹ out of a total riverine DIP export of 0.25 Tg P yr⁻¹). The contribution of agriculture to the total P load (including DIP, dissolved organic P, DOP and particulate P, PP) may however be larger, specifically for PP, which is about 40% of the total P export to waters in Europe (50% is DIP and 10% is DOP). The share of agriculture in PP export is determined by (i) surface runoff of P in particles of fertilizer and manure, (ii) agricultural practices (e.g. tillage) that affect the erosion rate and (iii) elevated P contents of soil material eroded from agricultural fields due to application of P fertilizers and animal manure. The effects of agriculture on PP export are likely to occur much faster than the strongly delayed effect of DIP export through the soil systems but these mixed contributions make it very hard to assess what the agricultural contribution to the PP load is.

Finally, we have used the MSA indicator as a measure for terrestrial biodiversity. MSA represents an index of the naturalness of an ecosystem. Compared with more traditional measures (e.g. monitoring species changes), this measure has two main advantages: it is possible to attribute biodiversity loss to certain sectors (PBL 2014), and the effect of alternative scenarios on biodiversity can be quantified (PBL 2012, 2014). The patterns of change indicated by the MSA are largely similar to those indicated by other measures as the living planet index as developed by WWF or red list indices (SCBD 2014). A limitation of the MSA indicator is that it does not yield comprehensive information on the actual distribution and abundance of species, such as the status of endangered or threatened endemic species.

While some uncertainty is associated with the individual emission estimates, other parameters might dominate the uncertainty of the estimated shares. For example, while the estimate of N₂O emission factors might be associated with an uncertainty of up to 50% (at the European scale), the uncertainty around the estimated share of crops that is used as feed determines the uncertainty of livestocks' contribution on total agricultural N₂O emissions. A bias in the total feed translates directly into a bias of the estimated share of emissions from livestock production, as it not only determines the share of crops used as feed, but also the amount of CH₄ emissions from enteric fermentation in ruminants and manure excretion (which is calculated on the basis of animal retention data) and consequent emissions from manure management. This value is obtained from statistical sources (market balances) and is further constrained by energy and nutrient requirement calculations for major livestock types.

No data are available whether farmers prefer domestically produced crops or imported crops for feed; therefore this value is highly uncertain. Because of the lack of information, we considered equal preference. However, this uncertainty concerns only primary crops that are used for both food and feed, which make only 12% of the total feed, while the rest comes from non-marketable crops (82%), such as grass, fodder maize and beet, or secondary crops (6%). Non-marketable crops are all domestically produced; secondary feed stuff is dominated by imported soya bean.

There are two main routes to reduce the environmental impacts of livestock production:

• (i)  
  technical measures (reduce emissions intensity/land use intensity and
• (ii)  
  lower livestock production in the EU with demand side measures, i.e., a reduction of food losses and wastes and/or dietary shifts.

Our study presents a 'status quo' analysis (attributional LCA) and does not examine emissions without (or with less) livestock production (case ii). What would happen if livestock production is reduced in Europe has been discussed in depth in Westhoek et al (2014, 2015). Based on the observation that the intake of protein as well as saturated fats by European inhabitants is far above the maximum recommended level (WHO 2007, Westhoek et al 2011), the authors showed that reducing the consumption of meat, dairy and eggs in the EU27 by 50% would lead to a decrease of N_r emissions by 40% and a reduction of GHG emissions by 25%–40% with expected substantial health benefits (Westhoek et al 2014, 2015). Those results hold for two contrasting scenarios on the use of the 'freed' land that would not anymore be required for feed production, i.e., a 'greening scenario' with enhanced production of bio-energy and a 'high price' scenario with increasing export of cereals. They can be regarded as conservative scenarios, as beneficial environmental effects outside Europe had not been quantified, such as the subsequent prevention of land conversion outside Europe (Stehfest et al 2013), or the reduction of GHG emissions from bio-energy production (or other options such as afforestation).

From the production side many technical, structural or policy mitigation options are being discussed, addressing feed production (e.g., precision agriculture and agronomic nitrogen use efficiencies), livestock production (e.g. grazing and feeding management and feed supplements, improved herd structures), or housing and manure management (Thornton and Herrero 2010, Gerber et al 2013, Golub et al 2013, USDA 2013, Cohn et al 2014, Havlík et al 2014, Hou et al 2014, Van Middelaar et al 2014, Winiwarter et al 2014, Van Doorslaer et al 2015). The benefits of sustainable extensification practices have also recently been explored for Europe (van Grinsven et al 2015). Bouraoui et al (2014) have shown that high reductions of nitrogen losses to water could be achieved in Europe by an optimized use of organic manure. Additional emission reductions could be achieved by decreasing the wastage of the supplied proteins (Westhoek et al 2011, Bellarby et al 2013, Grizzetti et al 2013).