Corresponding Author: K. Senthilkumar, UMR 1220 TCEM, INRA, 71 av. Edouard Bourlaux, BP-81, F-33883 Villenave d'Ornon, France. (firstname.lastname@example.org)
 Global biogeochemical cycles have been deeply modified by human activities in recent decades. But detailed studies analyzing the influence of current economic and social organizations on global biogeochemical cycles within a system perspective are still required. Country level offers a relevant scale for assessing nutrient management and identifying key driving forces and possible leaks in the nutrient cycle. Conceptual modeling helps to quantify nutrient flows within the country and we developed such an approach for France. France is a typical Western European country with intensive agriculture, trade and an affluent diet, all of which may increase internal and external P flows. Phosphorus (P) was taken as a case study because phosphate rock is a non-renewable resource which future availability is becoming increasingly bleak. A conceptual model of major P flows at the country scale was designed. France was divided into agriculture, industry, domestic, import and export sectors, and each of these sectors was further divided into compartments. A total of 25 internal and eight external P flows were identified and quantified on a yearly basis for a period of 16 years (from 1990 to 2006) in order to understand long-term P flows. All the P flows were quantified using the substance flow analysis principle. The results showed that the industrial sector remained the largest contributor to P flows in France, followed by the agriculture and domestic sectors. Soil P balance was positive. However, a positive P balance of 18 kg P ha−1 in 1990 was reduced to 4 kg P ha−1 in 2006, mainly due to the reduced application of inorganic P fertilizer. The overall country scale P balance was positive, whereas half of this additional P was lost to the environment mainly through the landfilling of municipal and industrial waste, disposal of treated wastewater from which P was partially removed, and P losses from agricultural soils though erosion and leaching. Consequences for global P resources and soil and water compartments are discussed. Some opportunities to more effectively close the P cycle in France by both improving the intensity of P recycling and decreasing losses are quantified.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The global biogeochemical cycles of nutrients have been deeply transformed in the past decades [Galloway et al., 2008], leading to aquatic and terrestrial ecosystem eutrophication [Walker et al., 2009]. Human activities appear to be the main driving forces of such transformations [Bennett et al., 2001]. But detailed studies analyzing the influence of current economic and social organizations on global biogeochemical cycles are still required [Kaye et al., 2006]. Such studies require a systemic perspective due to the many interactions within societies. For example, fertilizer use depends on the organization and intensity of agriculture, whereas flows between agriculture and domestic sectors depend on the organization of the food and feed industry. In other words, agriculture is closely linked to other economic sectors such as food production, domestic food consumption, trade organization and waste recycling. Understanding nutrient dynamics at the global scale thus implies taking different economic sectors and their interactions into account. This can be done using a Substance Flow Analysis (SFA) approach where the flows of a particular substance into, out of and through a geographically defined area are quantified [Villalba et al., 2008].
 Country levels offer a relevant scale for assessing nutrient management within a global perspective and with SFA approach [Liu et al., 2004; Antikainen et al., 2005; Mishima et al., 2010]. Detailed and homogenous data are more likely to be available at such a scale. Moreover, it offers a clear link with public policies. Finally, country level studies offer opportunities for analyzing interactions between different economic sectors. As a consequence, it may be a relevant scale for identifying the key forces underlying nutrient use today and opportunities for better nutrient resource management in the future. We developed such an approach taking France as a case study. France is a typical Western European country with intensive agriculture, trade and affluent dietary habits. We therefore hypothesized that heavy P flows leading to heavy P consumption and depletion could be observed. We used current data that was as detailed as possible for material flow in order to identify possible losses along the food chain.
 We developed such an approach for Phosphorus (P). P is a major nutrient for life of all higher plants and animals, and it is not substitutable. However, maintaining high levels of P bioavailability in agricultural soils by using P fertilizers increases the risk of phosphorus transport to surface water [Heathwaite et al., 2003]. Phosphorus is the element that determines the ecological status of most inland waters, since it is the substance that usually limits biological growth in fresh water. Moreover, the main source of P, phosphate rock, is a non-renewable resource and its future availability is becoming increasingly bleak [Cordell et al., 2009]. Currently, 178.5 Mt of phosphate rock, equivalent to 23 Mt of P, are mined every year. Ninety percent of it is used for food production as fertilizer (approx 82%) and feed additives (approx 7%) and the remaining is used as detergents [International Fertilizer Industry Association (IFA), 2011]. Considering the predicted 50–100% increase in P demand by 2050 as a result of increased global demand for food and changing diets, available rock phosphate resources will be depleted in another 50 to 200 years [Sibbesen and Runge-Metzger, 1995; Steen, 1998; Smil, 2000; Van Vuuren et al., 2010].
 The objectives of this study were: (1) to identify and quantify P flows, stocks and balances across and within different sectors in France; (2) to quantify the French P efficiency of crops and animals, as well as P recovery and recycling from waste management systems and to identify opportunities for improvement; and (3) to understand how closed the P flows in France are and their consequences for soil and water compartments.
2. Materials and Methods
2.1. System Design
 The conceptual design of P stocks and flows at the country scale are presented in Figure 1. The geographical boundaries of the system analyzed are the land borders of France (excluding forests, islands and overseas territories). The system was divided into subsystems including agriculture, industry, domestic, environment, imports and exports, and the subsystems were again subdivided into compartments (Figure 1). The subsystem compartments are linked by P inflows and outflows where the outflow of one compartment becomes the inflow of another. Atmospheric deposition and imports are considered as P inflows into the system, and exports and losses are considered as outflows. P stocks and flows were quantified on a yearly basis from the year 1990 to 2006 to identify the long-term changes in P flows, whereas a five-year average was calculated for the years 2002–2006 to quantify the annual P stocks, flows and balances. Natural ecosystems such as forests were not considered in this study for several reasons. First, their contribution to national P flows is low. For instance, the losses to water bodies through runoff and leaching under forest soils are often insignificant (<0.1 kg P ha−1 y−1, even in clear-cutting conditions [Yanai, 1991; Ranger et al., 2007]). Moreover, the P content of wood is very low, around 0.1 kg P t−1 of wood [Augusto et al., 2000; André et al., 2010], leading to an estimation of only 1–2 kt P y−1 for French national wood production. Second, the interactions between these ecosystems and the system under study were very low: natural ecosystems do not receive mineral or organic fertilizer and their resulting products (e.g., paper and cardboard), besides representing minor P flows, were not accounted in this study.
2.2. Data Sources and Quantification Methods
 Data concerning the quantity of material flow in and out of each compartment were collected from European and French databases (European statistics (EUROSTAT); Ministère de l'Agriculture, de l'Alimentation, de la Pêche, de la Ruralité et de l'Aménagement du Territoire (AGRESTE); Food and Agricultural Organisation (FAO); Union des Industries de la Fertilisation (UNIFA); Institut Français de l'Environnement (IFEN); Agence de l'Environnement et de la Maîtrise de l'Energie (ADEME); Syndicat des Industries Françaises des COproduits (SIFCO)). The P concentrations of all the flow materials were collected from existing literature. Flows were quantified by multiplying the material flow with their P contents. Some flows were cross-checked by comparing their values using alternative calculation methods and data sources. All the quantified annual P stocks, flows and balances were expressed in kt P y−1. The Substance Flow Analysis (SFA) method was used to quantify all the P flows in this study. The basic objective of the SFA method is to quantify the pathways of a substance in, out and through a geographically considered system. The basic steps of SFA are (i) the design of a conceptual model of the studied system in terms of compartments and flows; (ii) the collection of data about material stocks and flows corresponding to the conceptual model and related nutrient content of these materials; and (iii) the quantification of nutrient stocks and flows within the system. SFA was used previously to calculate the P flows at global [Liu et al., 2008; Smit et al., 2009], national [Liu et al., 2004; Antikainen et al., 2005; Mishima et al., 2010] and regional [Senthilkumar et al., 2012] scales.
2.2.1. P Stocks
 Phosphorus stocks in soils, living animals, food and feed storage and the human population were quantified at the end of each year. P stock in soil was calculated by multiplying the agricultural area by a global average total P content of 3.75 t P ha−1 in the top 50-cm soil layer [Smil, 2000], which is well within the limit values of 1.8 to 5 t P ha−1 in French agricultural soils [Némery and Garnier, 2007]. P stock in living animals was calculated by multiplying the live weight of all living animals with their P concentration. P stock in humans was estimated by multiplying the human population by an average P content of 0.6 kg person−1 [Smil, 2000]. Food and feed P stock was calculated by multiplying the quantity of food and feed stock with its respective P concentration. P stock in the ‘crop’ compartment was considered to be zero since all the harvested crops were allocated to other compartments in the form of crop residue, fodder to animals and crop products for processing.
2.2.2. P Flows
 Twenty-five internal and eight external (import and export) P flows were identified and quantified. The internal flows are identified by the flow numbers 1–25 and the external flows are identified as i1–i4 for imports and e1–e4 for exports, respectively (Figure 1). A detailed description of each flow is presented in Table 1. The methodology used to quantify each flow and its data sources is presented in Table 2. The list of flows that are cross-checked is presented in Table 3 together with their alternative quantification method and data source. It was difficult to accurately quantify P losses through soil erosion and leaching at the country scale because of the non-availability of an average total P content of eroded agricultural soils in France. However, we used a 0.5 kg P t−1 of soil data to calculate erosion losses [Liu et al., 2008] and a P leaching rate of 0.1 kg P ha−1 y−1 to calculate leaching losses (Table 2) [Némery et al., 2005].
Table 1. Description of the P Flows Identified and Quantified in This Study
Quantity of P transported through harvested crops, including both economic produce and crop residues. All arable crops and home gardens were included, while forest and public gardens were excluded.
Quantity of P returned to the soil through crop residues, either directly or after being used in animal production.
Quantity of P in fodder produced to feed animals in France.
Quantity of P excreted through manure (urine + litter) by all animals in a year. The number of standing animals in a year was assumed to be static throughout the year and all the P excreted in manure is returned to agricultural soils.
Quantity of P from harvested production of crops other than fodder crops that are used for food and feed and industrial use (including biofuel).
Quantity of P through slaughtered animals, milk and egg production.
Quantity of P carried through the seed material used in all arable crop production.
Quantity of P from indigenously produced and imported feed fed to all animals in France. Feed input includes all cereals, pulses, root crops, oilseeds and oilcakes, vegetables, fruits, feed phosphates and mineral phosphorus.
Quantity of P used for fertilizer application for crop production in France.
Quantity of P in food products (plants, animals and seafood) consumed by humans.
Quantity of P from textile products including sanitary textiles and pet food.
Quantity of P through waste from the food processing industry.
Quantity of P in industrial and animal waste that is reused as pet food and animal feed.
Quantity of P through the municipal waste generated by households and small industries.
Quantity of P from waste composted and applied to soil. This waste was collected as municipal waste and waste from food processing industries including slaughterhouses.
Quantity of P from waste landfilled directly or after incineration that was collected as municipal waste and food processing industrial waste including slaughterhouses.
Quantity of P flow from detergents used in-households and small industries.
Quantity of P from human waste (urine + feces) and municipal wastewater (gray + black water), including detergents and kitchen washwater.
Quantity of P from the portion of sludge produced in treatment plants that is composted and applied to agricultural soils. This also includes sludge produced by households with individual collection tanks.
Quantity of P from the portion of sludge produced in treatment plants that is incinerated and landfilled.
Quantity of P from treated wastewater that was discharged into a body of water.
Quantity of P from all fishery products including all catches from marine and inland waters.
Runoff and erosion
Quantity of P lost from agricultural soils to water bodies through the process of runoff and soil erosion from agricultural lands.
Quantity of P lost from agricultural soils to water bodies through the process of leaching from agricultural lands.
Quantity of P added to agricultural soils through the process of wet and dry deposition of atmospheric phosphorus.
Live animal import
Quantity of P from the import of live animals.
Food and feed import
Quantity of P from the import of food and feed products including mineral P feed.
Quantity of P from the import of phosphorus fertilizer (raw material and processed fertilizer).
Quantity of P from the import of detergents. We assumed that all the P used in detergents was imported.
Live animal export
Quantity of P from the export of live animals.
Food and feed export
Quantity of P from the export of food and feed products.
Quantity of P from the export of phosphorus fertilizer.
Quantity of P from the export of detergents.
Table 2. Equations and Data Sources Used to Calculate P Flows in the Analysisa
= Fodder requirement per animal × Number of animals
Number of animals: AGRESTE; Fodder requirement: 1, 2, 3
= P requirement per animal × Number of standing animals – animal production
Number of standing animals: AGRESTE; P requirement per animal: 4, 5, 6, 7, 8
= Actual food consumption + Organic municipal waste
Quantity of actual food consumption: 9; P content of food products: 10, 11; Quantity of organic municipal waste: EUROSTAT; P content of municipal waste: 12
2.2.3. P Balances
 The annual P balances were calculated at the country scale, subsystem scale and compartment scale. The overall country scale P balance was calculated by two ways. First by calculating the P balance of import and export alone and second by adding atmospheric deposition to import and environmental losses to export, respectively. The equations employed are as follows:
where: PB_France = country scale P balance for France (kt y−1), IMPORT = P inflow through imports (kt y−1), EXPORT = P outflow through exports (kt y−1), ATMOSPHERE = atmospheric deposition of P in agricultural soils in France (kt y−1), LOSSES = P losses to water bodies and landfills (kt y−1), I = number of imports, E = number of exports, L = number of losses.
 The P balances at the subsystem scale and compartment scale were calculated as:
where P_balance = P balances at subsystem and compartmental scale (kt y−1), INFLOW = P flow through inflow (kt y−1), OUTFLOW = P flow through outflow (kt y−1), I = number of inflows, O = number of outflows.
2.2.4. P Efficiency, P Recovery, P Recycling and P Losses to the Environment
 The annual country scale P efficiency of agricultural soils and animals was calculated for a 16-year period from 1990 to 2006. Soil and animal P efficiency were estimated as follows:
where: Crop_uptakeP = total P removed from soils through crop uptake (flow 1 in Figure 1) in kt y−1, Soil_inputP = total P input in agricultural soils (sum of flows 2, 4, 7, 9, 15, 19 and 25 in Figure 1) in kt y−1.
where: Animal_outputP = total P flow through animal outputs such as milk, egg and slaughtered animals (flow 6 in Figure 1) in kt y−1, Animal_inputP = total P inflow in animals through fodder and feed inputs (sum of flows 3 and 8 in Figure 1) in kt y−1.
 The quantity of P recovered from food and feed processing waste, municipal waste and wastewater was calculated for the years 1999–2006. The P recovery percentages were calculated as follows:
where: P_recoveryww = P recovered from wastewater (%), Sludgep = quantity of P recovered in sludge produced in treatment plants (kt P y−1), Wastewaterp = quantity of P flow through wastewater produced in France (kt P y−1).
where: P_recoverymw = P recovery from municipal waste (%), Recoveredmwp = quantity of P recovered from municipal waste, including both composted and incinerated fraction (kt P y−1), Municipal_wastep = quantity of P flow through the total organic municipal waste produced in France (kt P y−1). The municipal waste fractions ‘paper and cardboard’ and ‘composite waste’ are not included.
where: P_recoverypw = P recovery from food and feed processing waste (%), Recoveredpwp = quantity of P recovered from food and feed processing waste (kt P y−1). This included both composted, reused and incinerated fractions. Processing_wastep = quantity of P from the total food and feed processing waste generated in France (kt P y−1).
 The P ‘recovered’ is the total P collected (in case of municipal and processing waste) or extracted (in case of wastewater) from these wastes. The P ‘recycled’ is the portion of P inflow through waste that is actually recycled/reused to agricultural soils or feed processing, the remaining P being landfilled. The annual P losses to the environment (water bodies and landfills) from agricultural soils, waste processing industries and wastewater treatment were added together to determine the annual P losses in France.
 Average P flows and stocks at the country scale for France for the years 2002–2006 are presented in Figure 1.
3.1. P Stocks
 The total P stock for 29.6 M ha of agricultural soils in France is approximately 111,075 kt P. The P stocks for animals and humans are 47.2 and 39.8 kt P, respectively. The long-term calculation showed a decreasing trend in animal P stock and an increasing trend in human P stock as the total animal and human population decreased and increased over the years, respectively (data not presented). The quantity of P stored in food and feed stock is approximately 25 kt P, which fluctuated over the years in France (Figure 1).
3.2. P Flows
3.2.1. P Flows in Agriculture
 Overall, the agricultural sector received a total P inflow of 533 kt P y−1, with a total outflow of 380 kt P y−1 (excluding internal flows within the subsystem). Agricultural soils in France received a total of 778 kt P y−1. Application of animal manure itself contributed 40% of the total P input in agricultural soils followed by inorganic P fertilizer (37%), crop residues (14%). The remaining 9% came from the application of recycled wastewater sludge, municipal waste compost, atmospheric deposition and seed inputs.
 A total of 564 kt P was removed from agricultural soils annually through crop uptake, in which 112 kt P returned to the soil as crop residues. Of the remaining 452 kt P, 213 kt P were used by animals as fodder consumption. Permanent pasture and meadows contributed 59% of the P flow in fodder, followed by temporary pasture (21%) and fodder maize (15%). The remaining P in fodder was obtained from minor fodder crops such as lucerne, clover, fodder kale, fodder beet, etc. The balance of 239 kt P was contributed by cereals (75%), oilseeds (12%), pulses (5%), root crops (4%), vegetables (2%), fruits (0.8%) and industrial crops (0.4%). The biofuel crops were included in cereals and oilseeds. The corresponding crop products were earmarked for the food and feed processing industries (Figure 1) where they were processed mostly for human consumption, feed for animals or for export.
 The long-term P flows and balances in agricultural soils showed a drastic reduction of P inflow to soils, while the P outflow through crop uptake and losses to the environment through erosion and leaching were maintained over the years, resulting in a steady decline of soil P balance (Figure 2). The total P inflow in agricultural soils was 36 kg P ha−1 y−1 in 1990, which was reduced to 24 kg P ha−1 y−1 in 2006, while the P outflow was maintained at approximately 18–23 kg P ha−1 y−1. Although there was a reduction in P inflow, French agricultural soils maintained a positive P balance. However, the positive P balance of 18 kg P ha−1 y−1 in 1990 was reduced to 4 kg P ha−1 y−1 in 2006 (Figure 2).
 Long-term P flows in agricultural soils in France based on the type of flow material are presented in Figure 3. In the 1990s, relatively large quantities of inorganic P fertilizer were used when compared to the P inflow through animal manure and other inputs (crop residue, seed, compost, etc.) for agricultural production. This confirmed the high dependency on imported P fertilizers in the early 1990s (Figure 3). However, inorganic P use was dramatically reduced, making animal manure the predominant P source for agricultural production in recent years in France.
 A standing animal population of approximately 320 million in France consumed about 390 kt P y−1, in which 213 kt P y−1 were obtained through fodder and the remaining 177 kt P y−1 through feed input (Figure 1). Of the total P consumption per animal, nearly 80% was converted as manure and the remaining 20% as animal products such as milk, egg and slaughtered animals. Of the total manure production of 310 kt P y−1, 70% was produced by bovines, followed by swine (10%), poultry (10%), sheep and goat (7%), equidae (2%) and rabbits (2%).
3.2.2. P Flows in Industry
 Overall, the industrial sector received a total P inflow of 826 kt P y−1, with a total outflow of 809 kt P y−1. A total of 318 kt P y−1 was provided by inorganic P fertilizers, 78% of which was imported as manufactured fertilizer and the remaining 22% manufactured by importing raw materials such as phosphoric acid and rock phosphate. Nearly 90% of this inorganic P fertilizer was used in domestic crop production and the remaining 10% was exported.
 The food and feed processing industry received 471 kt P y−1 through raw agricultural products (Figure 1). Of this, 239 kt P y−1 was provided by crop products, 80 kt P y−1 by animal products, 113 kt P y−1 by imported food and feed and a small amount (2 kt P y−1) by seafood. The major P outflows from the food and feed processing industry were feed input to animals (177 kt P y−1), export (119 kt P y−1), food for domestic human consumption (78.5 kt P y−1), production of processing waste (30 kt P y−1) and processed goods such as textiles and pet food (7.7 kt P y−1). Distinguishing food and feed material before processing for specific purposes was difficult. As a result, the quantification of food waste recycled as feed was not possible.
 Around 70 kt P y−1 was provided by solid waste generated in France. Municipal waste contributed 57% of this flow and the remaining 43% was contributed by food and feed processing industry waste. Approximately 10 kt P y−1 of solid waste (mainly from slaughterhouse waste) was reused by the food and feed processing industry to produce pet food. Of the remaining 60 kt P y−1 of solid waste, 50% was earmarked for agricultural soils and the remaining half was landfilled either directly or after incineration. Phosphorus flow through the use of P-based detergents was 33.2 kt P y−1, and we assumed that all the P used to manufacture these detergents was either directly or indirectly imported into France.
3.2.3. P Flows in the Domestic Sector
 Overall, the domestic sector received a total P inflow of 119 kt y−1and a total outflow of 126 kt P y−1. An average of 119 kt P y−1 was used by the French human population, of which 66% was contributed by processed food available for consumption, 28% by use of P-based detergents and the remaining 6% by processed goods such as pet food. Half of the P in processed food ended up in the municipal waste in France, accounting for 40.4 kt P y−1 (containing both non-edible fractions such as peels and bones and lost fractions such as uneaten food). Considering the quantity of P contributed by processed food (78.5 kt P y−1) and the human population of France (around 60 million inhabitants), we estimated that the per capita P available for consumption was 1.24 kg P capita−1 y−1. Of this, 0.63 kg P capita−1 y−1 ended up as municipal waste and the balance of 0.61 kg P capita−1 y−1 was actually ingested by the French population. These results are well in line with the previous estimated annual P ingestion of 0.55 kg P capita−1 y−1 in different parts of the world [Centre Européen d'Etude des Polyphosphates, 1997; Cordell et al., 2009].
 The total quantity of P carried in wastewater produced in France was 85.5 kt P y−1. This included human P excretion, use of P-based detergents and other household wastewater. Using our estimates on actual P ingestion by the French population (0.61 kg P capita−1 y−1), we calculated the quantity of P excreted through human waste as 38.5 kt P y−1, considering the fact that almost all P ingested by humans is excreted as waste [Jönsson et al., 2004]. Human excretion contributed 45% of the P flow in wastewater, followed by the use of P-based detergents (39%). The remaining P might have come from kitchen wash and wastewater collected from small industries by treatment plants. The quantity of P recovered from wastewater was 40.3 kt P y−1. In the years 2002 to 2006, 59% of the sludge was applied to agricultural soils either directly or after composting, supplying 24 kt P y−1. The remaining 41% of the sludge was landfilled, accounting for 16.3 kt P y−1. After recovering some of the P in the sludge, the remaining P (45.2 kt P y−1) was earmarked as treated wastewater and eventually discharged into water bodies.
3.2.4. P Import and Export
 On average, 464 kt P were imported and 165 kt P were exported to and from France annually. Fertilizer imports contributed 69% of the total P imported, followed by imports of food and feed (24%) and detergents (7%). Among the food and feed products imported, mineral P feed contributed 42%, followed by oilcakes (39%), and the remaining P was imported through crop and animal products (Figure 4). Most of the P was exported through food and feed materials (80%), followed by fertilizer (18%). In contrast with food and feed imports, 92% of the P exported was in the form of crop products, and the remaining in the form of animal products and oil cakes (Figure 4).
3.2.5. P Balances and P Losses to the Environment
 The overall country scale P balance, calculated as a balance between import and export, declined from 689 kt P y−1 in 1990 to 263 kt P y−1 in 2006 (Figure 5a). When the atmospheric deposition as an inflow and environmental losses as an outflow were added to the calculation, the P balance was 568 kt P y−1 in 1990 and 157 kt P y−1 in 2006 (Figure 5b). In both cases, the country scale P balance was still positive.
 A total of 138 kt P y−1 was lost to the environment, 66% of which was discharged into water bodies and 34% of which was landfilled. The major contributors of P loss to water bodies were disposal of treated wastewater (45.2 kt P y−1), runoff and soil erosion (43.3 kt P y−1) and P loss through leaching (3 kt P y−1) from agricultural soils in France. Among the landfilled waste, 65% was in the form of solid waste and the remaining 35% in the form of incinerated sludge.
3.3. Cross-Checking of Results and Data Reliability
 Cross-checking of results using alternative methodologies and data sources was found to be effective in confirming our results. Fodder production was compared to fodder requirements of the standing animal population. Both variables were found to be on the same order of magnitude, thus confirming the accuracy of our calculation (Figure 6a). Similarly, animal excretion plus animal products were compared to animal P requirements. The results were on the same order of magnitude (Figure 6b), with a difference of just 4% between them. Finally, the quantity of P from processed food was compared to the P from actual food consumption and organic municipal waste. The results were on the same order of magnitude and thereby confirmed our results (Figure 6c).
3.4. P Efficiency and P Recycling
 Soil P efficiency increased from 0.47 in 1990 to 0.76 in 2006, while animal P efficiency remained unchanged at 0.2 during the same period (Figure 7). The percentage of P recycled through waste recycling was between 74 and 79% for waste generated from the processing industry, while it was 34–45% for municipal waste over different years in France (Figure 8). P recycled from municipal waste showed improvement over the years (from 1999 to 2006) as more municipal waste was composted and applied to soil.
 Though the quantity of P recovered in sludge production from wastewater was 43–49%, it was only 24–30% actually recycled during the study period (Figure 8). The remaining sludge was incinerated and landfilled. Although there was an increase in the quantity of wastewater produced and collected in France (from 4.5 billon m3 in 1998 to 8.2 billion m3 in 2006), the increase in the quantity of sludge produced in treatment plants was only from 0.98 million t in 1998 to 1.13 million t in 2006. However, the P content of the sludge increased due to the higher number of treatment plant station applying sludge tertiary treatment. This made the P recovery in wastewater more efficient. Based on our estimation, the average P content of the discharged treated wastewater from treatment plants was decreased from 9 mg P L−1 in 1998 to 5.5 mg P L−1 in 2006.
4.1. French Contribution to Global P Depletion
 In the year 1990, France imported 871 kt P y−1, of which 85% was imported as inorganic P fertilizer either in the form of manufactured or raw material. In the year 2006, the total import was reduced to 428 kt P y−1, of which only 67% was inorganic fertilizer. Therefore, France accounted for 3.2% of the global rock phosphate depletion in 1990 and for 1.2% in 2006, considering an annual quantity of rock phosphate mined of 23 Mt of P y−1 [IFA, 2011], whereas it represents only 0.9% of the global population. Additionally, France may also indirectly contribute to rock phosphate depletion through large imports of oilcakes from South America and, thus, increased fertilizer requirements in this region [Schipanski and Bennett, 2012]. On the other hand, fertilizer use in France also contributes to food and feed use in other countries since export represents 165 kt P y−1, mainly through crop products. This strengthens the need to understand the key forces underlying P fertilizer consumption at the country scale within a systemic perspective.
 The reduced use of inorganic P fertilizer observed in France from the year 1990 to 2006 is in line with what was observed in the whole Western Europe. However, at global scale the major inorganic P fertilizer consumers, expressed as a percentage of global use, are China (30%), India (15%), USA (11%) and Brazil (8%) and the global mean inorganic P fertilizer consumption was 13 kg P ha−1 y−1 in 2008 [Smit et al., 2009]. National SFA exercises may help to understand the drivers of such inorganic P fertilizer consumption. For instance, in France, P accumulation in agricultural soils, losses through runoff and erosion, high degree of food wastage and moderate P recycling from municipal waste and wastewater may be important drivers of inorganic P fertilizer consumption. Several P SFA at country scale have already been designed for The Netherlands [van Enk et al., 2011], USA [Suh and Yee, 2011], China [Liu et al., 2004], Japan [Mishima et al., 2010], Sweden [Neset et al., 2008], Finland [Antikainen et al., 2005]. They are helpful to go beyond global P SFA [Villalba et al., 2008] that do not account for country differences in agriculture and trade intensity, food habits, etc. However, the comparison of these studies (e.g., through meta-analysis) is difficult due to differences in conceptual modeling of P flows and stocks among countries.
 Inorganic P fertilizer consumption may also results from factors that are not accounted for in national SFA. For instance, regional specialization of farming systems such as crop or animal production, responsible for decreasing the number of mixed farming systems, might have enhanced the heterogeneity of soil P balance in France. This specialization hampers P recycling between crop and animal production regions and may generate extra P fertilizer consumption. Such a specialization is observed in most countries: the problem highlighted for France in this study is representative of a general trend at global scale [Schipanski and Bennett, 2012].
4.2. P Cycling Impact on Soil and Water Compartments
 The total P inflow in soils ranged from 36 kg P ha−1 y−1 in 1990 to 24 kg P ha−1 y−1 in 2006. This is well in line with previous estimate in the Seine watershed (74,000 km2) in France giving a P inflow in agricultural soils of 20–25 kg ha−1 y−1 in 2000 [Némery and Garnier, 2009]. Such a decrease of P inflow led to a positive P balance of 18 kg P ha−1 y−1 in 1990 that was reduced to 4 kg P ha−1 y−1 in 2006 (Figure 2). Similar reduction in soil P balances were reported from Japan, where agricultural soil P balances were reduced from 153 kg P ha−1 in 1985 to 105 kg P ha−1 in 2005 [Mishima et al., 2010]. However, the magnitude of soil P balances in Japan is very high compared to French soils. In China, the average P input and output were 28.9 and 14.2 kg P ha−1 y−1 respectively, with a P surplus of 14.7 kg P ha−1 y−1 in 2004 [Chen et al., 2008], while in India the P input and output were 16.3 and 9 kg P ha−1 y−1 respectively with a P surplus of 7.2 kg P ha−1 y−1 during the same period [Pathak et al., 2010]. In France, the mean national soil P surplus accumulated over the 17 years period of 1990–2006 amounted to 163 kg P ha−1 averaging 9.6 kg P ha−1 y−1 in this study. This estimate is comparable to the 8 kg P ha−1 y−1 of P surplus reported for EU27 [Richards and Dawson, 2008; Smit et al., 2009] and 15 kg P ha−1 y−1 of mean P surplus reported for UK agricultural soils [Withers et al., 2001] even if the later estimate was for the 1935–2000 period. Such positive P balance led to increase in soil P availability. For example, in Brittany, an intensive French animal production region, the median Dyer soil P test were 133, 148, 156 and 159 mg P kg−1 soil for the periods 1980–1985, 1990–1994, 1995–1999 and 2000–2003, respectively [Lemercier et al., 2008]. These results show that the soil P availability increased under positive P balance but tended to a plateau due to the declining P balance that was observed in recent years. As a consequence of past P balances, 18% of the French agricultural soil is considered as very high soil P availability, possibly leading to environmental issues such as P losses to water bodies and eutrophication [Follain et al., 2009].
 According to our data, a total of 91 kt P y−1 reached water bodies for the period 2002–2006, mostly from treated wastewater and soil runoff (Figure 1). For the same period, the French Ministry of Environment estimated that a total of 22 kt y−1 of dissolved P reaches the sea, excluding the contribution of the Rhine watershed, inland lake discharge and direct sea discharge from treatment plant stations [Dubois, 2011]. However, if one considers that only 50–70% of total P is under dissolved form in water bodies [Jarvie et al., 2010; Némery and Garnier, 2007] and that P retention within stream reservoir may represent 15–25% of total P inflow [Garnier et al., 2005; Némery and Garnier, 2007] the corrected corresponding total P discharge in water bodies reaches 53 kt P y−1. The difference with our estimation may be due to the discharge to inland lakes and through Rhine watershed and possibly overestimation of P losses to water bodies. Moreover, our results suggested a moderate improvement in P recovery from wastewater (43% in 1999 to 49% in 2006) and in consequence a decrease in soluble reactive phosphate in stream water. This is well in agreement with the 42% decrease of orthophosphate concentration in main rivers observed in France between 1998 and 2006 that is attributed to the improvement of wastewater treatment process [Dubois, 2009].
 Losses through runoff and leaching to water bodies are partially due to soil primary mineral weathering. P input through weathering was estimated as 2 Mt P y−1 [Smil, 2000] and 1.6 Mt P y−1 [Liu et al., 2008] for global cropland. For France, it may represent 10–12 kt P y−1 if weathering is estimated proportionally to France agricultural area compared to global agricultural area. However, weathering was not included as an input flow to agricultural soils in this study since P stocks in soils were calculated based on their total P content; therefore, weathering was an internal flow of the soil compartment. Losses to water bodies may also originate from natural ecosystems, such as forests, that might represent a background flow. However, P runoff and leaching under forest appear to be insignificant (<0.1 kg P ha−1 y−1) in the literature, even in clear-cutting conditions [Yanai, 1991; Ranger et al., 2007].
4.3. P Efficiency and Opportunities for Improvement
 P management at the country scale is a subject of considerable debate in France. Both soils and the environment acted as major sinks in the P cycle during the years studied. The amount of P that was lost annually to the environment represents half of the amount of P applied on soils as inorganic fertilizer. Importantly, 66% of these losses were due to the disposal of treated wastewater and incinerated sludge and waste landfill, and the remaining 34% to soil erosion and leaching. Moreover, only 45 and 49% of P in municipal waste and wastewater respectively were recovered in the year 2006 in France. In contrast, 95% of P recovery from wastewater was reported in Finland through tertiary treatment [Sokka et al., 2004; Antikainen et al., 2005]. If a similar percentage was applied to P recovery from municipal waste and wastewater, an additional 61.6 kt P could be recovered annually, representing 22% of current fertilizer use. However, recycling all of the recovered P may be difficult due to multiple and sometimes heavy pollution of waste products, e.g., trace elements and medicinal residues. For these reasons, only 30% of P in wastewater was actually recycled in the agricultural soil in the year 2006 (Figure 1).
 Approximately half of the P entering the domestic food system ended as waste during the years studied (containing both non-edible fractions such as peels and bones and lost fractions such as uneaten food). This is well in line with the results of Hall et al.  that calculated a 40% food waste in the U.S. However, this result is slightly greater than that of Quested and Johnson  who estimated an average food waste of 30% in the UK by direct measurement. Of this 30% of waste, 80% was considered as being avoidable. Applying this percentage to French conditions would lead to a food savings of 32 kt P y−1. As a first approximation, if we assume that fertilizer is applied proportionally to food production, this food savings would lead to a 41% reduction in current fertilizer use.
 Soil P efficiency increased from 0.47 in the early 1990s to 0.76 in 2006, mainly due to the reduced application of inorganic fertilizer. During the same period, crop productivity was maintained or even increased in France (e.g., cereal grain production was increased from 55,060 kt in 1990 to 61,708 kt in 2006). Due to the reduced consumption of inorganic P fertilizer and increased crop production, the soil P balance was reduced from 18 kg ha−1 in the early 1990s to 4 kg ha−1 in 2006. If this trend continues in the coming years, sustaining higher soil P efficiency and further improving it will be an agronomic challenge for French agriculture. It is possible that crop productivity will be reduced with continued negative soil P balances. However, this perspective is vague since the P stocks in French soils are high due to continuous soil P enrichment in previous decades.
4.4. Quality of Results and Uncertainties
 Several country P budget studies have calculated numerous flows using balance calculation [Liu et al., 2004; Antikainen et al., 2005]. On the contrary, in this study, the country scale P flows were calculated independently. Even though the P inflows and outflows were calculated independently, the P balances at the compartment scale corresponded well to the P flows of the compartment (Figure 9). A negative P balance of 3 kt P y−1 in the animal compartment can be attributed to the slight decreasing animal population in France. A positive P balance of 16 kt P y−1 in the food and feed compartment can be attributed to an increase in food and feed material storage in warehouses and wholesale and retail centers at any given point in time. Therefore, a relatively high consistency of results assessed by cross-checking was achieved, thus ensuring the accuracy of the calculations (Figure 6). However, cross-checking requires alternative data sources and calculation methods for each flow and compartments, limiting cross-checking to only three flows and compartments in this study. However, similar calculation methods and data sources were used for the other flows, possibly leading to the same degree of reliability for any other flows (except for runoff and erosion, leaching and atmospheric deposition, and perhaps waste and wastewater management due to fuzzy data sources).
 Uncertainties of the P flow calculations may originate from two sources. First, from the data on the quantity of material and second from the P concentrations of the material used in the calculation. Most data on the quantity of material flow were obtained from the European and French databases where only a single value of each material flow is presented. Therefore, associated uncertainties could not be calculated. They are probably low since these databases were official databases used for several purposes, which means that several cross-checking procedures should have been performed before publication. However, uncertainties may arise due to the differences in P concentrations of the flow material. We used the most reliable P concentrations of the flow material which were calculated specifically for French conditions [Comité Français d'Étude et de Développement de la Fertilisation Raisonneé (COMIFER), 2007; Institut National de la Recherche Agronomique (INRA), 2007]. Moreover, at large spatial scales, such as the country scale in this study, it is meaningless to use extreme values of P concentrations and we stick to the average values from the literature. Some of the flow calculations are prone to some degree of uncertainties. For example, P flow from runoff and erosion was calculated by using an average soil P content of 0.5 kg P per ton of eroded soil which is well in line with the mean total P content in arable soils reported for the Marne river basin of France [Garnier et al., 2005]. However, the total P content may vary from 0.24 to 0.67 kg P per ton of eroded soil in France [Gavalda et al., 2005; Némery and Garnier, 2007] which may result 52% lower or 25% higher P losses through erosion and runoff but these are extreme values and cannot be extended at larger national scales.
 France high dependency on imported mineral P fertilizers for agricultural production in the early 1990s has been reduced by half in recent years. This has resulted in a lower positive P balance of agricultural soils. Sustaining or further improving agricultural production with reduced dependency on imported inorganic fertilizers will be an agronomic challenge for French agriculture in the coming years. Moreover, there is a need to decrease P losses to water bodies and landfill.
 This study leads the way to new perspectives. First, the current country scale soil P balance of 4 kg ha−1 may mask regional heterogeneity of soil P balances. Hence, the extension of P budget frameworks at lower scales would be useful. This would help to assess to what extent regional agricultural specialization may be an obstacle to efficient nutrient recycling and, thus, may be a key force underlying fertilizer use. Second, some opportunities exist at the country scale to better recycle nutrients. However, assessing such opportunities requires the design of consistent scenarios and process-based modeling based on the framework we proposed in this paper. Such a model is still to be developed.
 We are grateful to Philippe Eveillard and Martin Parmentier of UNIFA for providing us with fertilizer consumption data for France. We thank Chantal Gascuel and Riina Antikainen for providing us with insight to calculate some of the P flows in this study. We also thank our colleagues Monique Linères, Pascal Denoroy, Christian Morel, Laurence Denaix and Lionel Jordan-Meille for their constructive comments on this research and for helping us with data collection. This work was funded by Bordeaux Sciences Agro, Université de Bordeaux, and INRA, Department of Environment and Agronomy.