An Integrated Approach to a Nitrogen Use Efﬁciency (NUE) Indicator for the Food Production–Consumption Chain

: Reducing nitrogen pollution across the food chain requires the use of clear and comprehensive indicators to track and manage losses. The challenge is to derive an easy-to-use robust nitrogen use efﬁciency (NUE) indicator for entire food systems to help support policy development, monitor progress and inform consumers. Based on a comparison of four approaches to NUE (life cycle analysis, nitrogen footprint, nitrogen budget, and environmental impact assessment), we propose an indicator for broader application at the national scale: The whole food chain (NUE FC ), which is deﬁned as the ratio of the protein (expressed as nitrogen) available for human consumption to the (newly ﬁxed and imported) nitrogen input to the food system. The NUE FC was calculated for a set of European countries between 1980 and 2011. A large variation in NUE FC was observed within countries in Europe, ranging from 10% in Ireland to 40% in Italy in 2008. The NUE FC can be used to identify factors that inﬂuence it (e.g., the share of biological nitrogen ﬁxation (BNF) in new nitrogen, the imported and exported products and the consumption), which can be used to propose potential improvements on the national scale.


Introduction
Nitrogen (N) is a key nutrient, vital for the survival of humans and all other living organisms. While di-nitrogen gas (N 2 ) is abundant in the atmosphere, most organisms are unable to use this chemically unreactive form. First it must be converted or "fixed" into a reactive form such as ammonia (NH 3 ) or nitrogen oxide (NO x ). This and other forms of reactive nitrogen (all forms of nitrogen other than the inert atmospheric N 2 gas; N r ) are comparatively scarce and represent a limiting resource in most ecosystems and in farmlands (e.g., [1] and references therein).
By the end of the 19th century, the natural sources of fixed nitrogen were not sufficient for the food production needs of a rapidly increasing human population in Western Europe. The development and adoption of the Haber-Bosch process enabled widespread production and use of synthetic N-fertilizers, system. This figure also shows where losses (noted as "N-loss", not being reused) occur that reduce the efficiency. Up to the 'Food industry' box, the interactions are as represented by EU Nitrogen Expert Panel [8] in the agricultural system.
There are several methods available to determine the NUEFC. These methods have different aims, system bounds, limitations and advantages when used as a basis for deriving the full-chain NUE. We provide a review of these methods and indicators for NUE with the aim of proposing one (or a more coherent set of) indicator(s) for NUE in the whole food system in Europe (EU28) and its constituent countries, including an assessment of the advantages and disadvantages, the limitations, data availability and potential use.
The aims are to provide: (i) A literature review of existing indicators for full-chain NUE in the food production-consumption chain at the national and/or regional scales in European countries; (ii) A proposal for one indicator or a coherent set of indicators for NUE in the food system in Europe at the national or regional levels, which is 'linked' to the general concept agreed upon during the first meeting of the panel (EU Nitrogen Expert Panel, 2014), and which can be used by policymakers and practitioners in Europe (industry, consumers, NGOs, policy, research); and (iii) A demonstration of the use of the proposed indicator using European national data sets and a discussion of its usability.

Literature Review: Approaches to Estimating NUE
Four approaches are commonly used to assess nitrogen use efficiencies of food productionconsumption and/or the environmental impact of N use in food production: life cycle analysis, nitrogen footprint analysis, nitrogen budget analysis, and environmental impact assessment. These approaches are briefly described below, with more detailed information given in the Supplementary Material (SM).

Protein consumption
Atmospheric deposition Export Figure 1. The major components of the full-chain nitrogen use efficiency (NUE FC ) in the national food system. This figure also shows where losses (noted as "N-loss", not being reused) occur that reduce the efficiency. Up to the 'Food industry' box, the interactions are as represented by EU Nitrogen Expert Panel [8] in the agricultural system.
There are several methods available to determine the NUE FC . These methods have different aims, system bounds, limitations and advantages when used as a basis for deriving the full-chain NUE. We provide a review of these methods and indicators for NUE with the aim of proposing one (or a more coherent set of) indicator(s) for NUE in the whole food system in Europe (EU28) and its constituent countries, including an assessment of the advantages and disadvantages, the limitations, data availability and potential use.
The aims are to provide: (i) A literature review of existing indicators for full-chain NUE in the food production-consumption chain at the national and/or regional scales in European countries; (ii) A proposal for one indicator or a coherent set of indicators for NUE in the food system in Europe at the national or regional levels, which is 'linked' to the general concept agreed upon during the first meeting of the panel (EU Nitrogen Expert Panel, 2014), and which can be used by policymakers and practitioners in Europe (industry, consumers, NGOs, policy, research); and (iii) A demonstration of the use of the proposed indicator using European national data sets and a discussion of its usability.

Literature Review: Approaches to Estimating NUE
Four approaches are commonly used to assess nitrogen use efficiencies of food production-consumption and/or the environmental impact of N use in food production: life cycle analysis, nitrogen footprint analysis, nitrogen budget analysis, and environmental impact assessment. These approaches are briefly described below, with more detailed information given in the Supplementary Material (SM). The four approaches distinguished are not equally specific for estimating NUE, but they do provide information about the N use and possible environmental impacts of the N use. The four approaches are: (i) Life Cycle Analysis (LCA): A technique to assess the potential environmental and human health impacts associated with a product, process, or service by: (1) compiling an inventory of relevant energy and material inputs and environmental releases and (2) evaluating the potential environmental impacts associated with the identified inputs and releases (e.g., [9][10][11][12][13][14][15][16][17][18] (See Appendix ??, Table A1). (ii) Nitrogen Footprint: Total N losses to the environment resulting from the production of a defined unit of food (e.g., [19][20][21][22][23][24][25] (See Appendix ??, Table A2). (iii) Nitrogen Budget: The inputs and outputs of nitrogen across the boundaries of a system.
Can contain information about internal nitrogen fluxes within the system (e.g., [1,[26][27][28][29][30][31][32][33][34][35][36][37]). The farm-gate nitrogen budget is a common indicator for assessing the total inputs and outputs across a farm's boundaries. The nitrogen balance indicator measures the difference between the nitrogen available to an agricultural system and the nitrogen harvested and exported from the system in agricultural products. The food system waste/loss indicator: this category has received special attention by studies that estimate the loss of N through waste [38][39][40] (See Appendix ??, Table A3). (iv) Environmental Impact Assessment: A process of evaluating the likely environmental impacts of a proposed project or development, taking into account inter-related socio-economic, cultural, and human health impacts both beneficial and adverse (e.g., [41][42][43] (See Appendix ??, Table A4).
A literature review was conducted of the aforementioned approaches, and a total of 36 peer-reviewed studies were identified and categorized according to approach, country, and product as displayed in Appendix ??. The different studies cover 17 products and 30 countries. Of the papers collected, 11 used the LCA approach, six used the N footprint approach, 16 used the N budget approach, and three used the EIA approach.
All four approaches are useful for the purpose they serve. The level of detail varies and depends on the goal of the study as well as the availability of data. However, the N budget analysis approach is most suitable for estimating NUE FC ; this approach allows the direct estimation/derivation of NUE, N outputs and N surplus.

Literature Values
There were several studies that calculated the NUE FC for a country or a region for all food production using the budget method: World: country ranges from 1% (Mongolia) and 4% (Australia) to 106% (Nepal) and 112% (Rwanda), where nitrogen mining takes place ( [7]).
These studies show that the overall full-chain N efficiency (NUE FC ) is between 5 and 15%, which is a fairly small range. This means that on average, only 5-15% of the nitrogen invested in global food production is contained in final consumed food products. In some regions, the estimated NUE FC was very high (even more than 100%; Sutton et al. [7]), probably due to soil N mining.

Selection of Approach
The four full-chain approaches have strengths and weaknesses (Table 1). These strengths and weaknesses form the basis for our selection of a recommended approach for estimating full chain NUE. Other criteria include the feasibility of linking to the farm-scale NUE indicator, the system boundaries, and the current and foreseen data availability. The system boundaries defined here are set by the food system. This means that the energy and transport part of the food chain will not be included. However, the atmospheric deposition resulting from these N emission sources is included, similar to the study by the EU Nitrogen Expert Panel [8].
The LCA approach provides an established framework for detailed calculation of the N flows and losses to the environment in the different steps of the food system. Detailed data are needed for this approach. Although LCA has been completed for some food products in Europe, it is clear that the data needed for all European countries and products will not be available in the near future for a complete analysis of the whole food chain for all countries. Recently, Leip et al. [32] used the LCA approach together with the CAPRI (Common Agricultural Policy Regionalised Impact) model to estimate the N losses and impacts of livestock in Europe.
The N footprint and N budget approaches both use N input and output balances. The N footprint approach also considers N losses at each stage of the food production process, whereas the N budget looks at overall inputs and outputs for a given system. Given this, the N footprint approach is generally more data-intensive than the N budget approach. In addition, the N budget approach can be more easily applied at multiple scales (e.g., a step of the food system, an individual product, farm, region, country) given its flexibility in assigning the system bounds.
LCA and EIA both integrate environmental impact assessment; the other approaches simply assess N emissions to the environment. Impacts are region and climate specific; comparison of food systems based on impacts is therefore difficult. Furthermore, directly linking N use to the impact of N use is not part of the objective of 'how to estimate NUE of the food chain'.
In summary, most approaches for estimating N use in the food chain assess the losses of N from the food system and some link these losses to the environmental impacts (life cycle analysis, N-Footprint, environmental impact assessment). The N budget approach, by contrast, compares the overall inputs and outputs of a system. The focus on losses for the step-by-step approaches (LCA, N-Footprint, EIA) is less useful for calculating NUE, because NUE is based on the balance between N supply and the N in a final product. Furthermore, the different approaches should provide comparable results when applied to the same system since N supply should be the same as (the total N input-total N loss), provided all the N losses are known (including denitrification).
The N-budgeting approach is most useful for estimating the whole food system NUE, because it aligns better with the application objectives and the data required for its calculation are readily available. Nitrogen budget approaches compile the inputs and outputs of nitrogen across the boundaries of a system in order to model and quantify internal nitrogen fluxes by using a mass-balance approach. This approach captures all N in all the relevant steps throughout the food system and it can be applied to a range of products and locations with readily available international data sets. Table 1. Advantages and disadvantages of four different full-chain nitrogen use efficiency approaches: life cycle analysis, nitrogen footprint, nitrogen budget, and environmental impact assessment.

Definition of a Full-Chain NUE (NUE FC )
A food system can be regarded as a chain of sectors and activities: the producers, the collectors, the processors, the distributors, the retailers and the consumers. Within each sector, activities reduce the amount of N that makes it to the next stage of food production, e.g., through food waste or other losses to the environment. The result is that the total amount of N embodied in the product decreases with each step. For a particular food system, a distinction is often made between the production of plant and animal protein (e.g., [20]). Animal protein production adds other steps to the cycle and introduces the recycling of manure. Since nutrition is the essence of food, to achieve a maximum NUE FC , the Consumed N should be as high as possible and the New N as low as possible. The NUE for each step in the food chain ( Figure 1) is defined as Consumed N (the outputs) divided by New N (the inputs). For each step in each 'sector', a ratio can be derived, which represents the efficiency of the sector in terms of N use, e.g., Edible crop N/Crop N, which represents the efficiency of the processing of crops.
In order to calculate the NUE for the whole food system using the budgeting approach, the use of N in the system has to be estimated. The amount of N of each sector ( Figure 1) is needed to calculate the associated NUE of that sector. One complicating aspect is that NUE is influenced by the import and export of food and feed. These trade flows have to be included in the equation, as NUE FC represents the NUE of the entire food system. NUE FC is defined as follows: NUE FC = N food availability fertilizer + BNF + atm. dep + (import − export) + changes in stock (1) Here 'N food availability' (consumption) is usually determined by the N supplied to the households.
This equation includes N inputs to agricultural systems via the atmospheric deposition of non-agricultural sources (fossil-fuel-based emissions) and not the total N deposition and not the entire country (total surface area). The major components of NUE FC are shown in Figure 1. The net import (import-export) was used instead of placing 'export' in the denominator and 'import' in the nominator, because for countries where import and export are very high (throughput) this influences the efficiency. In the ideal case, the NUE for the imported products should be taken into account, but because of lack of data this is not included. The same holds for the exported products. In Equation (1) 'change in stock' is the annual net balance of the imports and exports of a country, which includes the storage of products. This term is zero over the long-term, but can be significant for a single year. Since we are focussing here on the national level, we assume that internal cycling of nitrogen within the country does not affect the NUE when using country-scale production and food availability.
The equation above defines the NUE over the whole food system, but it can be split into different parts, such as food processing, the agricultural system, or the consumer whenever data are available. NUE is represented by the agricultural system NUE as defined by the EU Nitrogen Expert Panel [8].
The data needed for the country level are listed in Table 2. With the data available in Europe, the country level can be calculated using the FAO Food Balance Sheets for protein consumption (availability of food), while for all commodities the N content was taken from [31] to obtain the total food N supply (N Consumed). Further, fertilizer statistics may be obtained from Fertilizers Europe (through the European Statistical Office-EUROSTAT), biological nitrogen fixation (BNF) from [46], or more recently from [31], and atmospheric deposition to agricultural lands from the European Monitoring and Evaluation Programme (EMEP) [47]. In addition, Edible crop N, Vegetable N and Consumed N for crops are needed, and data for animal protein food availability, which are available from the statistical database of the Food and Agriculture Organization (FAOSTAT). Furthermore, data on the import and export of feed, plant and animal protein are available from the FAO statistics.

Application of the Country Approach
The previous section provided the general definition and methodology for estimating NUE FC . In this section, the applicability of the methodology is tested using detailed national data for the Netherlands. This detailed result can then be compared to the simpler calculation with internationally available data sets in the following section. The N flows in the food chain of the Netherlands for 2005 are shown in Figure 2. The estimation of the NUE of the food chain (NUE FC ) is complicated because of the large trade in food and feed with other countries, which is all included in the FAO database.

Application of the Country Approach
The previous section provided the general definition and methodology for estimating NUEFC. In this section, the applicability of the methodology is tested using detailed national data for the Netherlands. This detailed result can then be compared to the simpler calculation with internationally available data sets in the following section. The N flows in the food chain of the Netherlands for 2005 are shown in Figure 2. The estimation of the NUE of the food chain (NUEFC) is complicated because of the large trade in food and feed with other countries, which is all included in the FAO database. Using the budget approach, the NUEFC in the Netherlands for 2005 was estimated at 18% [48]. The total food availability for consumption was 70 Kton N (50 Kton animal protein and 20 Kton vegetal protein). The net import was estimated at 40 Kton (523 Kton of import of mainly feed for animals and 484 Using the budget approach, the NUE FC in the Netherlands for 2005 was estimated at 18% [48]. The total food availability for consumption was 70 Kton N (50 Kton animal protein and 20 Kton vegetal protein). The net import was estimated at 40 Kton (523 Kton of import of mainly feed for animals and 484 Kton of exported food products). The total input of new N to the system includes fertilizer (280 Kton), imported feed (350 Kton), BNF (15 Kton) and atmospheric deposition to agricultural soils (55 Kton) [48]. Most of the atmospheric deposition results from the national sources (ammonia emissions from agriculture and NOx emissions from fossil fuels). The NUE FC can be split into a food processing and a food production component. Food processing was calculated using Equation (1) and the data from Figure 2; the efficiency of the food processing sector was 48% (including recycling of 15 Kton), which means that 48% of N is retained during processing. The production of food products (up to the food processing: the farm gate efficiency) has an efficiency of 36%.
The NUE FC for the consumption of plant and animal protein can only be calculated separately if the input as New N for the two sectors (crop production and livestock production) is known. Here, corrected fertilizer and manure inputs were used. This was achieved using the ratio of the output (80 Kton) over the input to livestock (260 Kton) to correct for the animal and plant protein inputs. The resulting NUE FC for plant protein in the Netherlands is 12% and for animal protein 6% (without the net export of protein).
The Netherlands has a high export/import ratio because of the large trade in feed and food. These large flows greatly affect the accuracy of NUE FC . Therefore, the net import (import-export) instead of the ratio of export over import is used in Equation (1).

NUE in the Netherlands Based on FAOSTAT Data Compared to National Statistics
For the application of Equation (1), consistent data have to be used, such as the FAOSTAT Food Balances (see Table 2). The level of detail used in the calculations depends on the availability of data. For each food category of the FAOSTAT food balance the nitrogen content was derived from [31] to calculate the nitrogen entering the household (N Consumed). FAOSTAT data were used on country N input (fertilizer, BNF, feed, import-export and stock) and protein food availability for consumption data to calculate the country average NUE FC using Equation (1). Table 3 shows the comparison between the Netherlands NUE FC as derived from the detailed national statistics using Equation (1) ( [48]) and when data from broader and more general databases available for all countries are used. The FAOSTAT data are different from the data for the Netherlands as derived from national statistics. The reason is mainly because different food components are included with different N contents. The NUE FC based on the FAOSTAT data is 18% higher than that based on national statistics. In recent years, manure export has increased from about 10 Kton/yr in the 1990s to currently 40 Kton/yr, which makes it an important component to take into account. However, these data are not available in FAOSTAT.  Table 3 shows that the NUE FC using the different datasets come to somewhat different outcomes (respectively 25% and 18%). When manure export is taken into account the results are within 20% of each other. The major differences are caused by the differences in imports and exports and in food availability data. The latter depends strongly on the components that are taken into account.

Country Level NUE FC in the EU Based on FAOSTAT
National data were used for a set of European countries to calculate NUE FC using Equation (1) and the data as indicated in Table 2. The resulting NUE FC for the countries are plotted in Figure 3 for the year 2008. It shows that there are large differences between countries: Ireland had the lowest NUE (10%), while Italy had the highest NUE (40%) in 2008. The difference in NUE for the countries can be explained by the difference in N input (such as BNF) and by the import-export ratios. In the next sections, first the NUE values will be compared to the literature values, followed by an estimate of the trends in NUE FC .
Sustainability 2018, 10, x FOR PEER REVIEW 10 of 31 Table 3 shows that the NUEFC using the different datasets come to somewhat different outcomes (respectively 25% and 18%). When manure export is taken into account the results are within 20% of each other. The major differences are caused by the differences in imports and exports and in food availability data. The latter depends strongly on the components that are taken into account.

Country Level NUEFC in the EU Based on FAOSTAT
National data were used for a set of European countries to calculate NUEFC using Equation (1) and the data as indicated in Table 2. The resulting NUEFC for the countries are plotted in Figure 3 for the year 2008. It shows that there are large differences between countries: Ireland had the lowest NUE (10%), while Italy had the highest NUE (40%) in 2008. The difference in NUE for the countries can be explained by the difference in N input (such as BNF) and by the import-export ratios. In the next sections, first the NUE values will be compared to the literature values, followed by an estimate of the trends in NUEFC.  [7] (Our Nutrient World, blue bars); by [34] in green; and data and methods proposed in this report for a country-level NUE indicator for 2008 in red.
Earlier FAOSTAT data were used by Sutton et al [7] on country N input and protein food availability to calculate country average NUEFC, where the total amount of N in plant and animal food is considered as a fraction of the total new N input from mineral fertilizers, BNF and imported feed and food. Figure 3 for the year 2008 shows a comparison of the data in Sutton et al. (2013), and those calculated here for some countries in Europe for the year 2008. Overall, the data by Sutton et al. are at least 25% lower than calculated here, mainly because of the net import considered here in order to correct for the share of inputs that is used for exported food and feed. Recently, Godinot et al. [34] used several methods to calculate NUE for agricultural production (not the food chain NUE). Their estimates, averaged for the years 2000-2008, are also plotted in Figure 3 for comparison. By definition the NUE for agriculture by Godinot et al., should be higher than the NUEFC since more losses further in the food chain after the farm gate reduce the overall NUE. Figure 3 shows that for half of the countries this is indeed the case, but for Italy, the Netherlands, Portugal, Sweden and the UK, the NUEFC values are higher than those for agriculture. The differences clearly represent the uncertainty in the data and approaches. If the differences between the three methods are considered as a measure of the uncertainty, the NUEFC calculated here has an uncertainty of at least 25% (calculated as the average standard deviation between the different national values). . NUE FC calculated by [7] (Our Nutrient World, blue bars); by [34] in green; and data and methods proposed in this report for a country-level NUE indicator for 2008 in red.

Country Level NUEFC in the EU between 1980 and 2011
Earlier FAOSTAT data were used by Sutton et al [7] on country N input and protein food availability to calculate country average NUE FC , where the total amount of N in plant and animal food is considered as a fraction of the total new N input from mineral fertilizers, BNF and imported feed and food. Figure 3 for the year 2008 shows a comparison of the data in Sutton et al. (2013), and those calculated here for some countries in Europe for the year 2008. Overall, the data by Sutton et al. are at least 25% lower than calculated here, mainly because of the net import considered here in order to correct for the share of inputs that is used for exported food and feed. Recently, Godinot et al. [34] used several methods to calculate NUE for agricultural production (not the food chain NUE). Their estimates, averaged for the years 2000-2008, are also plotted in Figure 3 for comparison. By definition the NUE for agriculture by Godinot et al., should be higher than the NUE FC since more losses further in the food chain after the farm gate reduce the overall NUE. Figure 3 shows that for half of the countries this is indeed the case, but for Italy, the Netherlands, Portugal, Sweden and the UK, the NUE FC values are higher than those for agriculture. The differences clearly represent the uncertainty in the data and approaches. If the differences between the three methods are considered as a measure of the uncertainty, the NUE FC calculated here has an uncertainty of at least 25% (calculated as the average standard deviation between the different national values). Figure 4 shows the trend for different countries in Europe between 1980 and 2011. Again, Italy shows the highest and Ireland the lowest NUE FC . Overall, NUE FC increased for all countries with time mainly due to the reduction in fertilizer use and a steady increase in the availability of protein for consumption, either produced nationally or through import. The changes over the years are not smooth because of the different changes in N-stock and import and export of N. In some years the stock increased and only reduced in the following year. Therefore, when trends are presented in NUE FC it is better to use moving averages of three years. Some countries showed an initial increase until 1990 and a subsequent decrease. This is especially true for the Eastern European countries and has to do with the economic situation that drastically changed since 1989 after the breakdown of the wall and with the reduction in fertilizer subsidies, which influences the input as well as the output.

Country Level NUE FC in the EU between 1980 and 2011
due to the reduction in fertilizer use and a steady increase in the availability of protein for consumption, either produced nationally or through import. The changes over the years are not smooth because of the different changes in N-stock and import and export of N. In some years the stock increased and only reduced in the following year. Therefore, when trends are presented in NUEFC it is better to use moving averages of three years. Some countries showed an initial increase until 1990 and a subsequent decrease. This is especially true for the Eastern European countries and has to do with the economic situation that drastically changed since 1989 after the breakdown of the wall and with the reduction in fertilizer subsidies, which influences the input as well as the output.

Factors Determining NUEFC
There are several factors that influence the NUEFC. In this section we discuss total N input, importexport and BNF. Five European countries were selected, which show different levels and trends, to determine the influences of the different factors: Hungary, Ireland, Italy, The Netherlands and Denmark. Figure 5 shows the relationship between NUEFC and the total input of new N (nominator of Equation (1)) on a per capita basis. Overall the trend is the same: if the total N input increases, NUEFC decreases. NUEFC strongly depends on the per capita N input, mainly because the total consumption increases with the population number. The curve in this graph is not unexpected, but the position and trends of the individual countries are interesting. These might be caused by the types of food produced in that country, but we do not have the information to test this. The aim should be to move up the curve towards higher NUEFC with lower per capita NUEFC.

Factors Determining NUE FC
There are several factors that influence the NUE FC . In this section we discuss total N input, import-export and BNF. Five European countries were selected, which show different levels and trends, to determine the influences of the different factors: Hungary, Ireland, Italy, The Netherlands and Denmark. Figure 5 shows the relationship between NUE FC and the total input of new N (nominator of Equation (1)) on a per capita basis. Overall the trend is the same: if the total N input increases, NUE FC decreases. NUE FC strongly depends on the per capita N input, mainly because the total consumption increases with the population number. The curve in this graph is not unexpected, but the position and trends of the individual countries are interesting. These might be caused by the types of food produced in that country, but we do not have the information to test this. The aim should be to move up the curve towards higher NUE FC with lower per capita NUE FC . Another important factor is BNF. In the calculation, BNF is taken as new N and thus if BNF increases, NUE decreases. However, countries that have a high NUE, such as Italy, have a higher BNF input. This might be explained by the type of agricultural system in the different counties. If BNF is high, the use of soil nitrogen for crop production is higher and less fertilizer is used, therewith the NUE of the agricultural system compared to systems using inorganic nitrogen inputs is lower. Therefore higher NUEFC is observed. This, however, needs to be further quantified.
As shown for the Netherlands, the import-export ratio is very important in the calculation of NUE. The net import was taken into account to calculate NUEFC instead of the ratio of export over import as Another important factor is BNF. In the calculation, BNF is taken as new N and thus if BNF increases, NUE decreases. However, countries that have a high NUE, such as Italy, have a higher BNF input. This might be explained by the type of agricultural system in the different counties. If BNF is high, the use of soil nitrogen for crop production is higher and less fertilizer is used, therewith the NUE of the agricultural system compared to systems using inorganic nitrogen inputs is lower. Therefore higher NUE FC is observed. This, however, needs to be further quantified.
As shown for the Netherlands, the import-export ratio is very important in the calculation of NUE. The net import was taken into account to calculate NUE FC instead of the ratio of export over import as explained earlier. For the Netherlands, the imported N, mainly as feed, is processed and exported as food leaving behind the wasted N. Italy depends on import for their food supply. For both countries it is shown that if the net import goes up NUE increases. There is no such relation for the net-exporting countries. It must be emphasised here that the NUE for the imported N has not been taken into account and thus is implicitly assumed to be 100%, just like the NUE for exported products, which is untrue. In the next stage of development of the NUE FC , either NUE of imported products should be included with a value that is based on the production and processing in the exporting country. Alternatively, one would have to follow in detail the process level of products being imported and exported, and attribute what happens in each (foreign) country individually-this is obviously a task not easily feasible, and needs to be tackled in a more elaborative approach (i.e., in the future).

Use of the Indicator
This section discusses if and how the indicator can be used to support policy and what is needed to further develop the indicator. A target for policy could be the use of the NUE indicator in relationship to per capita consumption. The World Health Organization recommends levels of protein consumption for age classes split between men and women. The average recommended level of consumption (expressed in terms of N) is 3 kg N/capita/year [49]. In general, it would be expected that the NUE will increase if the consumption per capita decreases, because the expectation is that this goes along with lower inputs. This holds especially for a decrease in processed food and a change towards more plant protein. For policy, it could be the aim to increase NUE, together with maintaining nutrition levels, but not exceeding recommended levels of protein consumption. For all countries the food availability for consumption is far above the recommended level. For most countries, the food availability goes up with higher NUE, except for Ireland, which shows no trend, and Hungary, which shows a decrease in food availability relative to NUE. There is therefore no clear relationship between food availability for consumption and NUE. This questions the usability of the indicator based on the current calculations. In the future this needs to be further tested.
Given the limited availability of data and the uncertainty associated with the different steps in the food system, it appears that the NUE FC indicator cannot be used easily for the optimization of the food production-consumption chain, nor can it be used for food labelling. The NUE FC indicator proposed here is suitable for supporting national policies and provides information about nitrogen consumption relative to the new N used to produce the consumed N. For food labelling, Leach et al. [20] propose to use the footprint approach.
The applicability of the NUE FC depends strongly on the availability of data, such as crop-specific fertilizer and manure data, data on the recycling of manure, BNF and deposition, soil N availability and use, N food availability for consumption, and loss fractions for each step in the food system. It was illustrated here for the Netherlands that when these data are available, the NUE FC can be calculated (excluding energy and transportation). It was shown that for the Netherlands the availability of protein N for consumption increased over the past decades both due to increased production nationally and through imports, while the N inputs decreased, implying an improved NUE FC . The current NUE FC is still in the low range, 12% for vegetal products and 6% for animal protein, and there is room for improvement of the methodology of estimating NUE FC , especially by addressing the balance of feed import.
The NUE FC indicator has the potential to provide information about the developments of the national NUE FC . For policy applications, one can imagine to set a so-called protein target, defined for example as the NUE for the whole food chain, or a food processing target. One could also think of targeting the share of plant protein in the whole food chain to increase NUE. However, the robustness of the indicator has to be improved by better including the NUE of imported protein and by separating national production from the import/export of protein. The ratio and the differences between animal protein and plant protein could also provide important information about the national NUE FC .

Next Steps
This initial study points in promising directions but, at the same time, raises numerous issues regarding data availability and the way that various values are used in calculations. The treatment of import and export of food is a particularly tricky aspect that has not yet been satisfactorily resolved. Estimations of biological nitrogen fixation also need better quantification. The way manure is treated in the calculations may also need further consideration. Deriving values for NUE FC at the national level is therefore challenging and differences between countries in the examples given should be interpreted with great care until the methods and data are further improved.
The framework presented in this paper does not take into account the role of the circular economy in N management. With the complexity of the food system, a circular economy is important in moving towards sustainability because it promotes the recycling of food wastes either as compost or feed to animals, recycling of sewage in agricultural soils, and reuse of crop residues either as compost or animal feed, etc. These recycled flows are not taken into account in this framework. If this indicator is to be used for policy and decision-making, the focus should be on how to stimulate the circularity in nitrogen management.
The logical next step is to further develop, improve and test a comprehensive NUE FC indicator as proposed now in Equation (1). This requires further data collection for Europe. This approach requires estimations of country-specific N flows for different steps in the food system, including the NUE for imported protein and the distinction between exported protein and national consumption. These N flows may be used to calculate the NUE FC for the new N inputs and include the recycling of crop residues and manure. This can be carried out for each country in Europe. An uncertainty analysis should be carried out on the national scale per crop and/or meat product. Also, the uncertainty of the data and the robustness of the inventory systems need to be assessed. Finally, the next steps needed to develop the indicators for relevant policy will include a test of the consistency and quality of the data and the set-up of a baseline calculation and a monitoring program.

Appendix A.1 Life Cycle Analysis (LCA)
Life cycle analysis quantifies the potential environmental impacts of a product, process, or service by assessing the inputs and outputs within a defined system boundary. The approach is conducted by first compiling an inventory of material inputs and environmental releases. The inputs and losses are then linked to environmental impacts, such as eutrophication potential or acidification potential. All losses and impacts are reported relative to a defined functional unit, such as 1 kg of beef or another product. Software tools are available for LCA.
A benefit of LCA is that it can be widely applied through available LCA software. Though LCAs can be applied to a range of 'products' and locations, their design also has limitations. When conducting a LCA, choosing an appropriate system boundary and functional unit is key to expressing results in a succinct yet widely applicable manner. Comparing LCA results is only appropriate when two systems have the same system bounds and functional units. Harris and Narayanaswamy [15] suggest choosing a functional unit that reflects the way a commodity is traded so that results may be translated across trans-national boundaries, and expanded to include economic analyses. Most studies examined here reflected this notion when choosing their functional units. Brentrup et al. [11] performed calculations in terms of tons of wheat, while [14] focused on 1 MJ of biofuel production. When using a software tool, the factors built in to the LCA tools are often not applicable to a system of interest and must be modified.
Though LCAs can link to environmental impacts, they are also data and resource intensive, which can be a significant disadvantage to regional-level analyses. Often data inputs are directly measured, which, though advantageous to site-specific investigations, are difficult to scale to a region.

Appendix A.2 Nitrogen Footprint
The footprint methodology captures total nitrogen losses to the environment resulting from the production of a defined unit of food. Usually completed on a regional scale, N footprints have the capacity to span the entire chain of a product from farm level and food processing to transport and consumption. Though often employing mass balance techniques that are commonly used in nitrogen budget approaches, N footprints aim to quantify commodities in a manner that allows inter-product comparisons to be made and conclusions to be drawn.
Though N footprints do not allow users to link products to environmental or social impacts, they do provide an advantage in that their outcomes are on the same platform, and can thus be compared. These comparisons are, in part, due to the ability of an N footprint to function on a regional level or above. Generalized processing and food waste data can often be used in place of site-specific measurements as are usually required in a LCA. N footprint approaches are also able to model internal N fluxes and credit the system where recycled N should be accounted for, as in [20]. Chatzimpiros and Barles [19] were able to account for BNF and atmospheric deposition in their N footprint calculations, while also crediting manure excretions back into the system for crop production.

Appendix A.3 Nitrogen Budget
Nitrogen budget approaches compile the inputs and outputs of nitrogen across the boundaries of a system in order to model and quantify internal nitrogen fluxes by using a mass balance approach. Within this approach, we classified two subcategories, the nitrogen balance indicator approach and the food system waste/loss approach. A nitrogen indicator uses the mass balance methodology to measure the difference between the N available to an agricultural system and the uptake of nitrogen by agriculture. The food system waste/loss category is similar, but emphasizes the losses of N through waste, especially during transport and processing, where losses can be high.
A major advantage to this approach is its ability to quantify indicators, which are often used to supply information on other variables that cannot be measured directly. In the case of a FAO study in 2011, Gustavsson et al. [40] used mass flows models to account for the loss or waste incurred during each step of the production process. This included agricultural production, postharvest production, handling and storage, processing, distribution, and consumption, indicating its widely applicable nature throughout each stage of the chain. However, though N budgets can effectively quantify intermediary steps, they often have difficulty determining the fates of the system outputs, and examining internal flows in a greater level of a detail.
Nitrogen budget studies: • Bleken and Bakken [26]: Nitrogen efficiency of food production in Norway. • Dalgaard et al. [27]: Poland, the Netherlands, France, Italy, Scotland, and Denmark (poultry, sheep, beef and dairy cattle, pigs, maize, forage, oilseed rape, "horticultural crops", silage maize, winter wheat, "leguminous plants", water buffalo-dairy, alfalfa, "vegetables", grass/clover, peas, oats, fava bean, rye, barley, triticale, and wheat) Cover inputs/exports: beet pulp, cereals, eggs, feed milk, fresh milk, alfalfa, grass, grass/clover, hay, various concentrates, meat, rape cake, soybeans, soybean oil cake, straw, sugar beets, whey, and silage-alfalfa, beet pulp, clover grass, grass, maize, whole crop. • Galloway and Cowling [28]: Global nitrogen efficiency of plant and animal production. • Grizzetti et al. [38]: quantifies the nitrogen loss to the environment related to food waste at consumption at the global and European scale and analyses its relative impact on the environment. Examined: cereals, roots and tubers, oilseed and pulses, fruit and vegetables, meat, fish and seafood, and milk (Global). • Gustavsson et al. [39]: Uses mass flows model to account for losses and waste incurred during each step of the commodity's food supply chain. Examined: cereals, roots and tubers, oilseeds and pulses, fruits, vegetables, meat, fish and seafood and dairy products. A variety of countries were considered, referred to in paper as industrialized (medium/high income) countries and developing (low income) countries. • Howarth et al. [29]: USA nitrogen budget, including food and energy production and consumption. • Isermann and Isermann [44]: Germany nitrogen balance for food and feed production and consumption. Westhoek et al. [45]: Assesses environmental impacts and health effects of diets with reduced meat and dairy in Europe.

Appendix A.4 Environmental Impact Assessment
Environmental impact assessments provide a means of evaluating the likely environmental impacts of a proposed product, while also taking into account inter-related socio-economic, cultural, and human health impacts both beneficial and adverse.
Though results are can be attributed to specific impacts, the analysis stops short here. Impacts are unable to be objectively compared, as a definitive ranking system has yet to be developed that will allow scientists to compare a product's carcinogenic effects to the degree of eutrophication it causes. Similarly to a LCA, it can be data intensive and site specific, negating the ability to generalize these impacts to another similar product. For this reason, environmental impact assessments cannot be widely applied without significant time and resource inputs.
Environmental impact assessment studies: • Van der Werf and Petit [51]: The guide reviews 12 indicator-based methods spanning a variety of countries and agricultural topics. The relevant studies are as follows: Malaysia (cabbage), four undisclosed locations in Europe (energy crops), Philippines (rice), France (integrative farming animal/crop diversity), and Switzerland (farm pollution sources).

Relevant Gaps
Inconsistent allocation methods are the most notable gaps in each of the reviewed approaches. It is during this phase of the calculations that the most bias is introduced. ISO 14040 standards recommend system expansion and substitution whenever possible, but often these are not viable options due to resource and data limitations. Harris and Narayanaswamy (2009) review alternative methods of allocation and the merits associated with each, including mass weight and economic value. The overwhelming conclusion states that choice of allocation method is subjective, and there has yet to be data supporting one alternative method over the other.
The following gaps were identified in the reviewed methods: • There is inconsistency in accounting for exports as a portion of national production.

•
When performing an N-footprint, some studies chose to include atmospheric emissions and others did not. This can drastically change a food product's environmental footprint.
• If including atmospheric emissions, estimating emissions can often be difficult since emission factors (EF) can vary drastically depending on a variety of variables (i.e., soil type, amount of precipitation, amount of N fertilizer applied, etc.) • Additionally, emission factors must be included from the extraction of raw materials to make fertilizer, from all farm equipment on a farm site, as well as emissions from manure.
• Variability in accounting for biological N fixation. • Not many studies differentiate between new (e.g., synthetic fertilizer) vs. recycled (e.g., manure) N inputs. • Difficulty with and a relatively high amount of inconsistency in approaching crop rotations, which, most articles agree, is a limitation in calculating an accurate N footprint.

•
Animal product allocation at farm gate. There are a variety of methods for determining how much of the N-footprint should be allocated toward multiple products resulting from one animal (for example: butter, milk, meat, etc.), and it is difficult to choose the appropriate methodology.
• Accurately assessing nitrogen leaching. This usually needs to use a model, and in order to estimate leaching, most studies suggested examining farm dynamics on a relatively small scale. • Main limitation: Connecting the results of the N-footprint, LCA, etc. to actual environmental impacts. This is the number one limitation in all of the studies.

Appendix B
Detailed literature review of four full-chain nitrogen efficiency indicators: Life cycle analysis (Table A1), nitrogen footprint (Table A2), nitrogen budget (Table A3), and environmental impact assessment (Table A4). Environmental impact of different N fertilizer rates in winter wheat production.
Assess resource depletion and environmental impacts. Impact categories: land use, climate change, toxicity, acidification, eutrophication.
n/a Extraction of raw materials, production and transportation of inputs, all ag operations in field, application of fertilizer (0-288 kg N/ha).
Nutrient removal in grain and straw (30-212 kg N/ha), emissions due to energy consumption, volatilization, leaching.

Caffrey and Veal 2013
Literature Review Review inconsistent methodologies associated with co-products, regional and crop specific management techniques, temporal variations, spatial variations, and nonpoint emission sources.
1. Land use change-examines direct and indirect implications regionally and globally. Discusses leachate and volatilization models, the consequences and processes of livestock production that should be considered. 2. Enteric fermentation and manure handling. 3. Considerations when including aquaculture. 4. Recommends the consideration of economics when comparing systems to determine the best mitigation strategies.
n/a n/a n/a

Cederberg et al., 2000
National (Sweden) LCA comparing organic and conventional milk production in Sweden in terms of environmental impacts and land requirements.
Conventional milk production was found to have a larger nitrogen surplus than organic per unit area, but organic has a greater N surplus per unit milk.
n/a n/a n/a 1. Goals: generally compared the environmental impact of farming practices or types of fed. 2. Allocation of co-products: economic allocation used in the past, but studies of beef and dairy have shown this to increase uncertainty. Preference-system expansion, physical relationships/causality, composition and economic value. 3. Merits of foreground (input processes, farm processes and production processes) and background (mining and extraction, grain production, transport) data sources. 4. Argues for uncertainty calculations to be incorporated into LCA results.
n/a Varies by study: fertilizer N, soil organic and inorganic N, mineralization, manure, green manure.

Liao et al., 2014 Varies
Study provides an overview of aspects that need to be taken into account for improved modelling of Nr releases in the LCA of crop production.
1. On-site crop production must include the harvested portion of the crop and the soil with a changing depth down to the water table. 2. Nitrate, nitrous oxide and ammonia should be distinguished within the crop product system and b/w the crop product system and the ecosphere. 3. Stand-alone LCA studies of crop production and those coupled with process-based models should be based on a consistent spatial scale. 4. Fate of Nr losses in the ecosphere should be explicitly modelled in the life cycle impact assessment phase of the LCA of crop production. 1. Two Dutch milk production systems (organic-11 farms, and conventional-10 farms) were examined from cradle-to-farm-gate. 2. Animal manure, animals, milk, roughage/bedding material were outside of system boundary. 3. Excluded medicines, seeds, machinery, buildings, transport or processing of raw milk.
1. Integral environmental impact (land use, energy use, climate change, acidification, and eutrophication) and hotspots were identified. 2. Conventional: purchased concentrates were found to be the hotspot in off farm and total impact for all categories. 3. Organic: purchased concentrates and roughage were found to the hotspots in off farm impact. 4. Allocation of multifunctional processes was done on the basis of economic allocation. n/a Concentrate N content, electricity, diesel, methane emission (enteric fermentation & manure management).
Leaching of nitrate and phosphate (farm-gate balance approach), ammonia volatilization (manure in stable, manure storage, grazing, application of manure & fertilizers), nitrous oxide emission.   Estimation of the N efficiency of the food producing sector in Norway, including the overall production system as well as specific products. Milk and other animal produce (incl. meat, live animals, eggs and wool).

Galloway and Cowling 2002 Global
Overall N efficiency of plant and animal production, starting with the input of N to a crop field and ending with the final food product.
1. Global average NUE is 14% for plant products and 4% for animal products. 2. Most of N used in food production is lost to the environment.
Global NUE = 14% for plant products and 4% for animal products.
N fertilizer or other input applied to a crop field, feed.
Food product, N loss at each stage of production.   China N use efficiency for crop production (26%), animal production (11%), and the whole food chain (9%).
Fertilizer inputs, crop yields and areas, number of animals, consumer diets, harvested crop and animal product nutrient content, rate and content of animal excretion.
Regional N and P flows, use efficiencies, and emissions by type.

Oenema 2006 Varies
Reviews N input-output budgets and N losses in livestock farming systems.
1. Generally, N inputs and losses increase in the order grazing systems < mixed systems < land-less systems.
2. Difficulties of establishing N budgets arise from the tendency of N to dissipate into the wider environment in a variety of species, including gaseous N species. 3. Standardization of methodologies is required to allow comparison of budgets. 4. Improve utilization of animal manure as fertilizer and manure management in general.

Parfitt et al., 2010 Global
Review of literature on food waste throughout the supply chain in developing, transitional, and developed countries.
Food waste is highest at the immediate post-harvest level in developing countries and the post-consumer level in developed countries.
n/a n/a n/a Shindo et al., 2003;2006 Regional (East Asia) Report the N load for East Asian countries using a budget approach, including N from biological fixation, energy production, human waste, and farmland.
1. Food production contributed more than 90% of the nitrogen load in East Asia. Fossil fuel N was only significant in Japan and South Korea. 2. N load was reported by sector as the difference between N input and N uptake.
East Asia NUE has a large range.
Fertilizer consumption, food balance sheets.
Food production and consumption, NOx emissions by region, N losses to waterways, denitrification, organic matter accumulation.
Westhoek et al., 2014 Regional (Europe) The environmental impacts (nitrogen emissions, greenhouse gas emissions, and cropland) and health effects of 6 alternative diets with reduced meat & dairy were reviewed.
1. Nitrogen emissions could be reduced by 40% from reductions in meat and dairy intake. 2. A variety of data sets and models were used to assess nitrogen inputs and emissions across the US. Resources include FAO for dietary data and the GAINS model for livestock excretion rates and N emissions.
Current Europe NUE = 18%. NUE would increase to 41-47% with meat and dairy reduction scenarios.
Food produced, food exported, emissions to air, emissions to groundwater and surface waters. 1. Taylor et al., (1993) use a Farmer Sustainability Index (FSI)-farmer production practices yield a positive or negative score, which are summed. 2. Ecopoints: assign scores to farmer production practices and landscape maintenance (used to establish payment incentives). 3. Agro-ecological Indicators (AEI): reflect the impact of one production practice on ALL environmental components. 4. Multi-objective parameters: accounts for a set of ecological, economic and social objectives chosen to solve problems in the current system. 5. Environmental management for agriculture (EMA): computer based systems produces eco-ratings reflecting environmental performance by comparing actual farmer production practices and site-specific details. 6. Solagro diagnosis: performance levels for criteria, the number of production systems w/in the farm, diversity of crops grown, management of inputs and management of space. 7. LCA for farm management: identifies the main pollution sources and evaluates possible modifications of the farms or farming methods. 8. Indicators of farm sustainability: assigns scores to production practices and behaviours.
n/a Use of non-renewable energy & other non-renewable resources, land use, water use, nitrogen fertilizer use, pesticide use, waste utilization.
There are some general gaps and trends across the different approaches. The general gaps in the four approaches are the following (depending on the application whether relevant or not): • Inconsistency in accounting for exports as a portion of national production. • Variability in accounting for biological nitrogen fixation (BNF), total soil nitrogen, nitrogen sources, and nitrogen sinks.

•
Limited differentiation between new (e.g., synthetic fertilizer) versus recycled (e.g., manure) N inputs. • Inconsistency in approaches that address crop rotations, which in most cases is a limitation in calculating crop specific N use.

•
Animal and plant product allocation to different final products at the farm gate. There are a variety of methods for determining how much of the N use should be allocated toward multiple products.

•
For the approaches based on losses, some studies chose to include gaseous emissions to the atmosphere and others did not. Atmospheric nitrogen deposition is often lacking as well as denitrification. In these studies, N leaching estimation is uncertain and therewith the N outputs.

•
There are limited connections between the N use results and actual environmental impacts.