ARTICLE https://doi.org/10.1038/s41467-019-14242-7 OPEN The shift of phosphorus transfers in global fisheries and aquaculture Yuanyuan Huang 1,2*, Phillipe Ciais 1, Daniel S. Goll 1,3, Jordi Sardans4,5, Josep Peñuelas 4,5, Fabio Cresto-Aleina1 & Haicheng Zhang1,6 Global fish production (capture and aquaculture) has increased quickly, which has 1234567890():,; altered global flows of phosphorus (P). Here we show that in 2016, 2:043:09 −1 1:59 Tg P yr (mean and interquartile range) was applied in aquaculture to increase fish production; while 1:101:14 −1 was removed from aquatic systems by fish harvesting. Between 1950 and 1:04 Tg P yr 1986, P from fish production went from aquatic towards the land-human systems. This landward P peaked at 0.54 Tg P yr−1, representing a large but overlooked P flux that might benefit land activities under P scarcity. After 1986, the landward P flux decreased sig- nificantly, and became negative around 2004, meaning that humans spend more P to pro- duce fish than harvest P in fish capture. An idealized pathway to return to the balanced anthropogenic P flow would require the mean phosphorus use efficiency (the ratio of har- vested to input P) of aquaculture to be increased from a current value of 20% to at least 48% by 2050 — a big challenge. 1 Laboratoiredes Sciences du Climat et de l’Environnement, LSCE/IPSL, CEA-CNRS-UVSQ, Université Paris-Saclay, 91191 Gif-sur-Yvette, France. 2 Commonwealth Scientific and Industrial Research Organisation, Aspendale, 3195 Victoria, Australia. 3 Department of Geography, University of Augsburg, Augsburg, Germany. 4 CSIC, Global Ecology Unit CREAF-CSIC-UAB, 08913 Bellaterra, Catalonia, Spain. 5 CREAF, 08913 Cerdanyola del Vallès, Catalonia, Spain. 6 Department Geoscience, Environment and Society, Université Libre de Bruxelles, 1050 Bruxelles, Belgium. *email: yuanyuanhuang2011@gmail.com NATURE COMMUNICATIONS | (2020)11:355 | https://doi.org/10.1038/s41467-019-14242-7 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-14242-7 O ne big challenge humanity faces today is the phosphorus fish, intensities of culture practice (e.g., extensive, semi-, and (P) dilemma. Phosphorus (P) is an essential element for intensive), aquatic ecosystems (e.g., freshwater, brackish water, or all forms of life on Earth. The rapid rise of human marine), background environment (e.g., water chemistry), rearing demand for food has quadrupled P inputs into the biosphere facilities (e.g., ponds, cages, pens, tanks, or raceways), number of since the preindustrial time. These inputs have primarily occurred species (e.g., polyculture vs. monoculture), and the socioeconomic through mineral P fertilizer addition1,2. In the long term, status of fish farmers. As a result, a data-driven quantification of P scarcity may threaten global food production because the ore anthropogenic impacts of fishery production on large-scale P deposits from which P fertilizer is made are depleting and likely budgets is challenging. This study fills the gap through providing to be exhausted in the future1,3. On the other hand, today, a data-driven quantification of the net impact of the global fishery the distribution of P is highly uneven4, and regional surpluses of on P flows, improving basic understandings of P biogeochemical P in croplands, inland waters, and coastal seas are not cycles, and providing support to identify the P management uncommon1,4–8. Excessive P in inland and coastal waters has targets. We reconstruct global P budgets driven by fishery pro- been widely recognized as the dominant driver of eutrophication, duction by an extensive compilation of data coming from global which degrades water quality, decreases biodiversity, alters eco- fishery production databases (see the “Methods” section, Sup- system dynamics, and results in dead zones7–11. These wide- plementary Fig. 2), fish stoichiometry from a whole-body P con- spread adverse consequences of eutrophication triggered recent centration (including bone) dataset with 1164 records for 224 fish worries that the planetary boundary estimated to be a load of species (see Methods, and Supplementary Figs. 3 and 4), and 0.89 kg P yr−1 per capita or 6.2 Tg P yr−1 in total to the ocean aquaculture farm system-level PUE estimates for a representative (mostly from the land) would be exceeded, i.e., its safe global range of aquaculture farm types with 168 entries (see Methods, operating space would be transgressed12–14. Supplementary Fig. 5). In addition, we also compile a P-retention Food production is the largest driver of large-scale anthro- efficiency (PRE) database across controlled feeding experiments pogenic P release into aquatic ecosystems, and recent global-scale (Supplementary Fig. 6). We define PUE as P harvested via fish P budgeting has focused on agriculture6,15,16. In total, 82.4% of biomass divided by P input via feed and fertilizer. PUEs are phosphate fertilizers goes to cropland and pasture. The high rate calculated at the farm level. Fish feed includes commercially of manure and mineral P fertilizer application to agriculture manufactured compound aquafeeds with different additives, results in around a half of the P not being taken up by plants, farm-made feeds from crop–livestock products and/or by-pro- increasing the risk of P being transferred into aquatic ecosystems ducts, and fish with low economic value or by-products recycled through erosion, runoff, and leaching6,17–19. Phosphorus applied from aquaculture industry etc. Fertilizers cover inorganic che- to cultivated soils in livestock manure exceeds the global mineral mical fertilizers and manures from a variety of sources, such as P fertilizer use17, and a third of the P transferred into freshwater poultry, ruminant waste, swine waste, and human excreta. PRE is is attributable to the livestock sector18. Human waste processing the fraction of P that is recovered in harvested fish biomass per or disposal, and the use of detergents also release P from con- feed P intake from controlled feeding experiments (see Methods sumed food products into inland and coastal waters. As a result, for details). current mitigation strategies and technological innovations con- We show the global net anthropogenic P flow between centrate on recycling and better management of soil, crop, and land–human and aquatic systems through fisheries and aqua- livestock P flows, and improved rates of recovery P from culture (Fig. 1; see also Supplementary Fig. 1, Supplementary wastewater19,20. methods: Boundary conditions). Here aquatic systems include The global fishery is an overlooked food production subsector both marine and freshwater systems. The net P flow consists of that is critical in land and aquatic nutrient flows21–24. Finfish, two major gross fluxes (opposite in directions): P flows crustaceans, and mollusks, hereafter generalized as fish, con- extracted from aquatic ecosystems by fish harvest (P-harvest) tribute substantially to the global animal protein supply for and P input flows to aquatic systems in the form of feed and humans (~17% in 2013). The annual global fish harvest, including fertilizers (P-input). We use P-harvest minus P-input, that is, both capture and aquaculture, has increased from 19 Tg in 1950 P-net, to quantify the net P transfer between aquatic and to 169 Tg in 201625. Fish harvest returns nutrient to land–human land–human systems driven by the global demand for fish systems but is currently an underrepresented aspect of anthro- meat. When P-net is positive, P moves out of aquatic ecosys- pogenic P fluxes globally and regionally. Aquaculture has been tems and is a short-term gain for land–human systems. On a the fastest-growing sector of food production over the past dec- longer term, fish P waste after consumption or during fish food ade25 and is expected to expand further22,26. However, its processing can be dumped to landfills, recycled to livestock and dependence on wild fish and crop livestock for feeds, the use of croplands, or recycled back to rivers and returned to aquatic water and land resources, and other environmental impacts on systems. Conversely, when the balance is negative, it means that aquatic ecosystems has cast doubt on the environmental sus- humans add to aquatic ecosystems more P for fish production tainability of aquaculture21,22,27–30, along with growing sustain- than they retrieve in harvesting wild and cultivated fish. Annual ability concerns on other food production sectors. As with the P-harvest (mean and its range, 1950–2016) is estimated from husbandry of livestock, aquaculture production relies on external combining fish production with the whole-body P concentra- supply of P either directly through feeds (e.g., for carnivorous tion for different fish species (see Methods and caption to fish) or through fertilizers that enhance the primary productivity Fig. 2). Annual P-input is derived from P-harvest and aqua- of aquatic ecosystems (e.g., for herbivorous and omnivorous culture system-level PUE (see Methods and caption to Fig. 2). species). Phosphorus that is not harvested might end up in inland PRE is treated as a reference PUE given by current technology and coastal waters and result in eutrophication9,31,32. More than and the PUE level that the global fishery management could 90% of fish farming occurs in Asia where P-use efficiency (PUE) potentially reach. Tracking of the fate of harvested P, the defined as the ratio of harvested to input P for a given farming relative form of P losses in aquaculture (e.g., excrements, system is generally low. For example, aquaculture PUE ranges uneaten feed, or by-products), and the location of P losses (e.g., from 8.7% to 21.2% in China32, indicating large loss rates to the during transport, in hatcheries, or during fish processing) environment. through time are not the major focus of this study (see Fig. 1, The global fishery production system is highly diverse with Methods and Supplementary Methods for details) due to lim- respect to harvested fish species (528 in 2014)33, fates of harvested ited reliable global datasets. 2 NATURE COMMUNICATIONS | (2020)11:355 | https://doi.org/10.1038/s41467-019-14242-7 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-14242-7 ARTICLE interquartile range) in 1950 to 1:101:14 −1 in 2016; at the 1:04 Tg P yr Land-human same time, P-input (aquaculture feed and fertilizer) grew from 0:015 0:010:007 to 2:043:09 −1 (Fig. 2, Supplementary Figs. 7 and 1:59 Tg P yr Waste Vegetation 8). Our global estimate of P-input for 2010 (1.46 Tg P) inferred Human from farm-level PUE data is close to the value from the World- Livestock Fish database (1.11 Tg P) obtained by multiplying the estimated farming area by the parameter representing per unit area nutrient P rock P Fertilizer Fish feed Harvested and feed input28. P-harvest estimated here is the largest pathway to fish fish that transfers P from aquatic ecosystems to land, compared with currently known pathways, i.e., 0.0056 Tg P yr−1 from anadro- P-harvest mous (migratory) fish34, 0.099 Tg P yr−1 from seabird colonies35, P-input and 0.16 Tg P yr−1 from sea salt deposition36. In all, 99% of P-harvest came from wild capture (mostly marine capture) in Aquatic ecosystem 1950, and this share decreased to 62% in 2016 (Supplementary Figs. 7 and 8). Because wild fish capture dominated fish pro- duction in the early decades of the record (Supplementary Fig. 2), Wild fish P-harvest outweighed P-input, resulting in a net removal of Aquaculture P (positive P-net) from aquatic ecosystems, which reached a P-balance = P-harvest - P-input maximum at 0.54 Tg P yr−1 in 19861987 1970 . Aquaculture took off dramatically in the 1980s accompanied by increased P-input, Fig. 1 Overview of the global fishery P flows. Harvested fish transport P while PUE did not increase at the same rate as P-input. This from aquatic ecosystems to land (blue arrows, Methods: Data, P budget), expansion of aquaculture led to a net flux of P from land–human while fish farming requires P input primarily through fish feed and P systems to aquatic ecosystems—a negative P-net for the global fertilizer (orange arrows, Methods: Data, P budget). The global fishery P fishery sector. The turning point from a positive to a negative transfers involve complex interactions among terrestrial vegetation, global P-net occurred around 20042008 1997 . Today, the global P-net is livestock, and human society and waste managements (dashed gray clearly negative and amounts to 0:951:99 0:50 Tg P yr−1. In com- arrows, see Supplementary Fig. 1 for more details). We focus on the fishery- parison, leaching, runoff, and erosion losses of fertilizer P from caused major perturbation of P flows between aquatic ecosystems and croplands37 to freshwater are reported to be 0.6 Tg P yr−1 from land. Aquatic systems here include both marine and freshwater systems. Mekonnen and Hoekstra 37 over 2002–2010, and Lun et al.6 We adopt a land or human-centric viewpoint. We call external P that goes report a larger P loss rate of 3.7 Tg P yr−1 from croplands to directly into the aquatic environment through fisheries and aquaculture, P- water bodies through runoff over the same period. For refer- input, and P that moves out of the aquatic ecosystems through harvested ence4–22, Tg P yr−1 are transported from rivers to oceans (Ref. 38: fish, P-harvest. We use P-net that is the difference between P-harvest and 4 Tg P yr−1; Ref. 39: 9 Tg P yr−1; Ref. 40: 22 Tg P yr−1). The role P-input, to quantify the net P budget between aquatic ecosystems and land. of the global fishery sector in loading P from land to aquatic Details of the boundary conditions are provided in the Supplementary systems thus represents an important component of anthro- Information. Animal and tree silhouettes are from Microsoft PowerPoint pogenic P transfers. Freshwater aquaculture contributes to most (office 365) icon. (84–94%) of the P-input. The share of marine aquaculture increases through time (Supplementary Fig. 8). Within the 3.0 P-harvest P-net freshwater aquaculture, most of P-input ends up in raising finfish P-input IQR (95–100%), and the share from raising crustaceans slightly 2.5 0.5 2004 2008 increases with time, to reach 5.31% in 2010. In the marine 0.0 aquaculture, most of P-input (>90%) ends up in finfish aqua- P budget (Tg P yr–1) 1997 2.0 –0.5 –1.0 culture during 1950–1970; however, after 1990, around 50% of 1.5 –1.5 marine aquaculture P-input goes into raising crustacean species –2.0 (Supplementary Fig. 8). 1.0 1950 1960 1970 1980 1990 2000 2010 At the continental scale, Asia has been driving the global shift to a negative P-net, with P-net remaining positive in Europe, 0.5 America, Africa, and Oceania, where P-harvest continued to 0.0 exceed P-input (1950–2016, Fig. 3). P-input first outweighed 1950 1960 1970 1980 1990 2000 2010 P-harvest around year 1988 >2016 2008 1974 in South Asia, by 1990 1982 in >2016 >2016 Year East Asia, 2005 1985 in Southeast Asia, and 2016 1998 in West Asia. In the year 2016, Asia alone contributed a negative Fig. 2 Global fishery phosphorus balance from 1950 to 2016. The blue line P-net of −1.13 Tg P yr−1 and the rest of the world a small positive shows the average P-harvest flux from fish. The orange line represents the P-net of 0.18 Tg P yr−1. average P-input into aquatic ecosystems through aquaculture. The inset Country-wise, enhancements in P-input from developing figure with the green line is the net P-net, i.e., P-harvest minus P-input. Asian countries, and to a lesser extent, decreases in P-harvest Light-gray shading indicates interquartile ranges (IQR, the 75th and 25th from wild fish captures in some developed countries, contributed percentiles) calculated from 1000 estimates with randomly sampled fish P to the overall negative trend in P-net, which crossed the zero line concentration, fish biomass within 50% percentile uncertainty for P- downward after 20042008 1997 (Fig. 4, Supplementary Figs. 12, 13, and harvest, and culture-system level P-use efficiency for P-input. Units are Tg 14). Between the recent decade (2007–2016) and the earlier P yr−1. Source data are provided as a Source Data file. decade when P-net was approaching its positive maximum (1980s), P-input in aquaculture increased by a factor of nine. Results and discussion Developing Asian countries contributed most of the increased Historical and future quantifications. The global P-harvest P-input, with China accounting for 60% of the global increase, (wild + aquaculture) increased from 0:210:24 0:19 (mean and India 10%, Indonesia 6%, Vietnam 6%, Bangladesh 3%, and NATURE COMMUNICATIONS | (2020)11:355 | https://doi.org/10.1038/s41467-019-14242-7 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-14242-7 P budget (Tg P yr–1) a South Europe b North Europe c East Europe d West Europe 0.100 0.015 0.10 0.075 0.02 0.010 0.05 0.050 0.01 0.005 0.025 0.000 0.00 0.000 0.00 e North America f South and Central America g Africa h Oceania P budget (Tg P yr–1) 0.06 0.20 0.075 0.010 0.04 0.15 0.10 0.050 0.005 0.02 0.025 0.05 0.00 0.00 0.000 0.000 i Southeast Asia j South Asia k West Asia l East Asia P budget (Tg P yr ) –1 0.6 0.015 P-havest 0.4 1.5 P-input 0.4 0.010 1.0 0.2 0.2 0.005 0.5 0.0 0.0 0.000 0.0 1960 1980 2000 1960 1980 2000 1960 1980 2000 1960 1980 2000 Fig. 3 Continental-scale fishery phosphorus balance from 1950 to 2016. Continent designation is based on FishStatJ version and details are 3.04.625 provided in Supplementary Table 2. Gray shading indicates the interquartile ranges (IQR, the 75th and 25th percentiles). Units are Tg P yr−1. Note that the Soviet Union is assigned as part of East Europe before its collapse and countries are assigned into different continents according to their geo-location after the collapse. Source data are provided as a Source Data file. Thailand 2%. Reduction in P-harvest was the main contributor to crustacean species, and the upper values are 44% (95th percentile) reduction in P-net from a few countries such as Japan, Russia, for finfish and 24% for crustacean species. Adopting co-culture Chile, Denmark, and Canada. farming systems (finfish + crustacean) may also increase PUE, We then address the question of scenarios for the future by up to 61% according to one study42, but there is a large evolution of P-net and how the fish production sector could uncertainty in the PUE of those systems, with few data become P-neutral by year 2050. Wild fish production has available43,44 (Supplementary Discussion: Data pattern). stagnated for the past two decades and is unlikely to grow Upper values of PRE measured in controlled experiments are significantly by 2050; on the other hand, the “business as usual” also informative: they indicate a potential gain in PUE that could aquaculture production is projected to reach 140 Tg (or 2.3 times be achieved in (idealized) closed culture systems with no leakage. its 2010 level) in 2050 after accounting for fish supply and The overall median PRE is 37% based on 348 feeding experiments demand, health of wild fisheries, fish prices, population growth, conducted in controlled environments (e.g., closed tanks) GDP growth, and technological progress26,41. We first establish a (Supplementary Fig. 6), that is, 1.85 times larger than median baseline scenario to 2050 by assuming that wild fish capture PUE. The 75th percentile of PRE is 52% from finfish and 21% will be stable at its mean level of 2005–2014. We calculate that from crustacean species, respectively. The 95th percentile reaches P-input to aquaculture will grow up to 3.42 Tg P yr−1 to match 78% for finfish and 26% for crustacean species. Finfish dominated the fish demand increasing from the “business as usual” the farmed fish production and contributed to 87% of total P in aquaculture production projection from Waite et al.26 with harvested fish (mean, 1950–2016), while the share of crustacean PUE staying at its current level of 20% (see Methods). In this species grew from near zero in 1950 to 6% in 2016 due to the baseline scenario, P-net will remain negative, and the imbalance decline in cultivating mollusks (Supplementary Fig. 7). If we between land and aquatic P flows will grow by a factor of two to assume a PRE of 21% from raising mollusks and with 87% of reach −2 Tg P yr−1 by 2050 (Table 1). We then build an idealized harvested fish P from finfish, the 75th percentile of PREs would alternative scenario where a neutral P-net is set as a global target correspond to an aquaculture PUE of 48%. The most important by 2050. To meet this target, the weighted (by fish production) cultivated fish species, the carps, comprise ~40% of the total global average PUE will have to increase to ~48%. To assess if aquaculture production by weight45 and their 75th percentile PRE such a PUE increase from current low value to a high efficiency of reaches 57%. 48% could be achieved with current best practices, we analyze the upper range of PUE among diverse aquaculture production systems, and upper range of PRE data from feeding experiments. Implications for the large-scale P budget. Human perturbation The rationale is that PUE could be increased by transforming of the global P cycle has mobilized a large amount of P from practices in all aquaculture farms so that they could reach their phosphate rock into the hydrosphere. The total P load into currently maximum achievable value. aquatic ecosystems from crop livestock spans a large range among The PUE of individual farming system spans a wide range studies and is reported to be 4 Tg P yr−1 (fertilizer and manure (1–167%, Supplementary Fig. 5). Our PUE database covers six of sources only) in 2000 from Bouwman et al.46, 5 Tg P yr−1 the seven largest aquaculture P-input contributors, including (fertilizer and manure), or 13.5–25 Tg P yr−1 if land use change China (59% of P-input), India (10%), Vietnam (6%), Bangladesh was additionally accounted from Peñuelas et al.7 over 2005–2011, (3%), and Thailand (2%) but excluding Indonesia (5%). Thus, the 12.9 Tg P from Chen and Graedel47 in 2013, and 9.7 Tg P yr−1 global weighted average PUE (20% during 2005–2014) mostly from Lun et al.6 during 2002–2010. The global fishery P input reflects the PUE from China due to its dominant share. In China, into aquatic ecosystems reached 2.06 Tg P yr−1 in 2016, which is the current median PUE is 19% from finfish and 12% from significant despite smaller than P load through crop livestock. 4 NATURE COMMUNICATIONS | (2020)11:355 | https://doi.org/10.1038/s41467-019-14242-7 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-14242-7 ARTICLE a P-harvest log(mean2007 – 2016 – mean1980 –1989) 80 10–1 60 10–2 40 10–3 20 0 0 –10–3 –20 –10–2 –40 –60 –10–1 –150 –100 –50 0 50 100 150 b P-input log(mean2007 – 2016 – mean1980 –1989) –1 80 10 60 10–2 40 10–3 20 0 0 –10–3 –20 –10–2 –40 –60 –10–1 –150 –100 –50 0 50 100 150 c P-net log(mean2007 – 2016 – mean1980 –1989) –1 10 80 60 10–2 40 10–3 20 0 0 –10–3 –20 –10–2 –40 –10–1 –60 –150 –100 –50 0 50 100 150 Fig. 4 Changes in the global fishery P budget. (a) P-harvest, (b) P-input, and (c) P-net between the 1980s and the recent decade (2007–2016). Note the log scale on the color legend and that the units are in Tg P yr−1. Country borders are adapted from TM World Borders Dataset 0.3. Table 1 Projections of mean global fishery P budget in 2050. food and fertilizer for aquaculture. The global fishery P-input that is not accounted for in the crop–livestock P loads, such as through compound fish feeds and mineral fertilizers, is nontrivial. Wild Aquaculture P-harvest P-input P-net production production For example, the global total of commercially manufactured compound aquafeed is estimated to be 34.4–39.6 Tg in 201248. Unit Tg Tg Tg P Tg P Tg P An average of 1% P concentration in compound aquafeed cor- Mass 98.7 140.0 1.41 3.42 −2.01 responds to more than 0.3 Tg P yr−1 input into aquatic ecosys- tems. In addition, crop-livestock P that enters aquatic ecosystems Per unit food protein supply, P-net from fish-driven food pro- relies on complex interactions between local topography, vege- duction (capture fisheries and aquaculture) ranges from 0.019 tation cover, climate, and land–water connections that regulate (here positive means a landward P movement) to −0.052 g (P) the fraction of surplus P going through erosion, runoff, and per g (Protein), which is still smaller in magnitude than that from leaching37. Fish feeds and P fertilizers are commonly directly the crop–livestock system, −0.025 to −0.138 g (P) per g (Protein) dumped into receiving waters to support fish production, (Supplementary Discussion, Supplementary Table 1). If historical resulting in the fraction of P that ends up in the water (commonly fish P protein had been supplied by the crop–livestock used as a parameter to estimate P loads)37 being different from system instead, there would have been more P loaded into aquatic the crop–livestock system, especially for marine culture in coastal ecosystems. However, when we investigated aquaculture sepa- regions. Leaking P from aquaculture system that ends up in rately, P-net went from −0.292 to −0.310 g (P) per g (Protein) the ocean may contribute to feed wild fish. It is likely to be (Supplementary Table 1). Some P from crop livestock is used as marginal considering the large amount of P available in the ocean NATURE COMMUNICATIONS | (2020)11:355 | https://doi.org/10.1038/s41467-019-14242-7 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-14242-7 at a global scale but might be important locally. We acknowledge aquaculture (IMTA) and integrated aquaculture/agriculture that not all P makes its way to the ocean. For example, in some (IAA), use wastes from one species or one subsystem (e.g., ponds, up to half of the nutrients may end up in sediments that aquaculture, crop, and livestock) to serve as food or fertilizer for can be reused for agriculture49. In all, 84–94% of the aquaculture another. These systems incorporate ecological principles to P-input went into the freshwater and only a portion could ulti- recycle nutrients, and in theory, are self-sustainable. The low mately enter the ocean. Such internal P loading, if not managed PUE from crustacean species could potentially be improved properly, would result in the legacy P problem that might extend through dietary phytase supplementation or coculture with finfish poor water quality issues for decades after adopting good man- species42–44. agement practice50. The higher aquaculture P-input from our In practice, despite researches documenting the benefits of data-driven estimation compared with the modeling result of integrated rearing systems, a large fraction of nutrients is unused Bouwman et al. 46 and Bouwman et al.51 can partly be explained and they are not recycled31, as witnessed by the decline of by the fate of P after entering the aquatic ecosystems. For traditional IAA, especially in China where there are increasing example, Bouwman et al.46 quantified P release from ponds and concerns about environmental sustainability53. Modern IAA, or assumed that particulate P was not released from pond systems, modern ecological aquaculture54, is promising: it improves while our study quantified the total P-input into aquatic systems aquaculture PUE but requires effective management and modern that include ponds. technology to upgrade traditional semi-intensive practices. Ref. 53 Through a data-based global quantification of the under- reviewed possible ways of successfully combining traditional represented yet critical land–aquatic P fluxes, we found the largest ecological aquaculture with modern pellet-fed methods. With pathway that transferred P from aquatic ecosystems to land, modern ecological aquaculture, a PUE of 48% would become an which has important implications for the biogeochemical cycle of achievable target technologically, despite economically there is a P. In the 1950s, total global P fertilizer application amounted to long way to go. Improving PUE in aquaculture is urgent in around 3 Tg P yr−1, of which 1 Tg P yr−1 ended up in aquatic China55 and China is playing a leading role in restricting systems52. A net P transfer of 0.2 Tg P yr−1 in the 1950s from the aquaculture P-input by adopting strict regulations aiming for a global fishery is a relative significant contribution to return green growth of its aquaculture during the recent decade56. It is P from aquatic systems to land. Historically, human demand for possible to expect a raise in aquaculture PUE in China in the fish may indirectly alleviate P shortage on land as wastes future with its aquaculture practice gradually shifting toward associated with fish processing or consumption would serve as environmental sustainability57. Countries like India, Vietnam, crop fertilizer or feeds for livestock. Compared with current Bangladesh, and Thailand, which contribute significantly to known pathways that transfer P from aquatic to land systems aquaculture, yet with a lower PUE than China, need more (0.0056–0.1634–36 Tg P yr−1), fish-driven landward P flux was the support in adopting advanced feeding technology and modern biggest yet largely overlooked, at least in the 1980s. We also practice. revealed a shift of the net land–aquatic P transfers driven by We need to pay attention to the role of capture fisheries and human demand for fish. If no waste management practice is aquaculture in returning P back into the land–human systems. implemented to reuse P accumulated in aquaculture systems, a A wide spectrum of technologies exist that would allow P to be net global fishery flux of −0.95 Tg P in 2016 (or −2.01 Tg P in recovered from heterogeneous waste flows, including fish- 2050) is considerable. Regionally, the threat of crossing the processing waste, transportation losses, residential, commercial, planetary boundary is even stronger for Asian countries, and and institutional food waste, human excreta, sewage networks, this is without considering fish trading that may alter regional and landfills58. For example, treated fish-processing wastes are P fluxes. Future studies focus on the spatial heterogeneity of widely applied as animal feed and soil fertilizer59–61. In total, 70% fishery P fluxes, the costs, and who experiences that the impacts of fish is processed before delivering to the customer, resulting in would be helpful in clarifying environmental and ecological 20–80% of fish waste depending on final products and fish benefits or harms locally. species60,62. China’s fish-processing industry could produce 0.65 Tg of fishmeal and 0.16 Tg of fish oil60. Globally, 10% of fish P harvest was returned to aquaculture through fishmeal and fish Outlook for a global P-neutral fishery management. Fish have a oil, and another 4% was recycled as fishmeal and fish oil for higher feed conversion efficiency (e.g., compared with beef and poultry, swine, human supplements, and other sectors, leaving pork) and fish production’s carbon footprint is lower than other 0.84 Tg P yr−1 of harvested fish P that could potentially be further animal production sectors22. As a result, aquaculture production recovered and reused (see Supplementary Discussion for details) is expected to grow further to sustain future demand for food22; it in 2010. The proportion of fishmeal and fish oil used to feed follows that P-input is going to increase unless mitigation stra- farmed fish is decreasing, partly due to the temporal instability in tegies are implemented to improve aquaculture PUE. We could supply, improved feeding technology, and rising prices, and this not detect an improving trend in PUE in our dataset because of trend is likely to continue in the future63. That means more the lack of repeated measurements over time. Nevertheless, we harvested P can potentially be reused for enhancing other aspects are optimistic in aquaculture becoming more efficient with of the human food chain in the future. Recovering P from urine modern technology and management playing a more important and feces has a long history and is being advanced by numerous role in future. studies today64–67. Recovering P from wastewater has been a Technically, the difference between median versus upper PUE priority subject of research with a range of scientific and values (e.g., 61% from China) and optimal PRE values (52% or engineering advances being made68,69. Many other strategies 78% from finfish) suggests that it should be feasible with current focusing on food and waste management are also being put technology to increase the current global average PUE (~20%) into practice, e.g., the diversion of food waste from landfill up to the 48% target in 2050. This increase might be achieved to agricultural land by composting70 and the application of through optimizing feeding efficiency (e.g., modifying formulated decentralized P recovery systems71. These technologies diets, feeding frequency, and culture environment) and promot- are promising to recover most P transferred from the aquatic to ing low-impact production systems (e.g., recirculating aquacul- land–human systems through harvested fish in future. As an ture systems, biofloc technology, and integrated systems). example, Norway is at the top of human development index Integrated systems, such as the integrated multitrophic rankings and has a P use efficiency of 92% in fishery with high 6 NATURE COMMUNICATIONS | (2020)11:355 | https://doi.org/10.1038/s41467-019-14242-7 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-14242-7 ARTICLE utilization of fish scraps from land-based processes72. P reuse and plants’ contribution to the fishery P budget is unlikely to be substantial. Ref. 51 recovery are currently far from mainstream practice58. Practical showed that seaweed removed <0.01 Tg P yr−1 from aquatic ecosystems. In this study, we do not include aquatic plants. Aquatic mammal production ranges from issues such as costs and mismatches between locations may 0.19 to 1.11 Tg, which is also not included in our P budget. discourage P recovery practice. We need sustainable policies at FAO acknowledges its concerns on the quality of biomass production data, and the international, national, and local levels to develop efficient data reported by FAO on behalf of reporting countries are incomplete with under- P recycling strategies and sound management plans to motivate representation of nonindustrial (e.g., artisanal, subsistence, and recreational) and industrial and household-level practice. Although fish trading illegal fisheries75–77. Therefore, we use the reconstruction of wild marine catches from Sea Around Us to correct the FAO database for the underrepresented fish complicates country-level commitments, countries with high harvest. This decade-long catch reconstruction integrates collective efforts from P-harvest, from China, Indonesia, India, Peru, Norway, Russia, hundreds of experts and covers the whole globe. Ref. 78 showed that the United States, and Japan to Vietnam, have a crucial role to play in reconstruction from Sea Around Us is on average 53% higher than that given by leading the practice of recovery and reuse of harvested fish P. earlier FAO reports (1950–2010), including fish that are not landed, i.e., discarded fish. Discards refer to fish that are not retained on board during fishing operations The global fishery does not occur in isolation—the fishery and are returned to water bodies. As discarded fish do not reach land, we assume P dynamics are closely coupled with other P mass flows and is that discarded fish do not play a role in land–aquatic P transfers. FishStatJ 3.04.6 one part of the global P biogeochemical cycle. Cultivated fish wild fish production is scaled by landed fish from Sea Around Us at the country species increasingly depend on feed inputs from the level from 1950 to 2014 with the average scaling factor decreasing from 1.29 (1950) to 1.06 (2004). The scaling factor for year 2014 is applied to year 2015 and 2016 as crop–livestock system30,73, while wild capture fishmeal and fish Sea Around Us only reported data up to 2014. The problem of misaccounting of oil are traditionally used to feed livestock. Researches focusing on fish biomass is not significant in aquaculture, at least mariculture79, and we use food, biodiversity, climate change74, and land use27 have now the FAO dataset for aquaculture. Production is expressed as per unit live weight started to integrate wild capture fishery, aquaculture, and (wet weight). P is unevenly distributed in fish organs. The skeleton (bone or cartilage) agriculture in assessing sustainability challenges. We propose generally has a higher P concentration compared with muscles. Existing food that future studies should incorporate fishery into P assessment nutrition databases, such as the FAO/INFOODS global food composition database and mitigation strategies. As the case study in Norway72 shows, for fish and shellfish—version 1.0 (uFiSh1.0), report P concentration of edible P use from both within and cross-sectors of aquaculture, fisheries, portions of fish, but this does not adequately capture the whole-body P and agriculture is far from being optimized. We need integrated concentration. We compile data from the available literature that report whole- body P concentration of finfish, crustacean, and mollusk species. A total of management systems that simultaneously optimize PUE from 175 peer-reviewed studies were compiled in the database (Supplementary Data 1). multiple sectors. A global effort to optimize, integrate, and The entire database includes 262 records for wild and 902 records for raised fish manage multi-sectorial P dynamics is the path to a sustainable across diverse species (224 in total), environmental conditions, dietary treatments, blue revolution in aquatic systems. and ontogenetic stages. Among these records, 1088 entries document P concentration for finfish, 41 for crustaceans, and 35 for mollusks. The database reports whole-body P concentration in mass fraction, i.e., the Methods mass of P divided by the live mass of fish. When a specific study reported dry weight-based P concentration, we transfer P concentration into wet basis using fish Overview. Our ability to quantify large-scale fishery P is restricted by the limited amount of data available. We compile large datasets on the whole-body P con- moisture content. If the study does not report fish moisture content, the average moisture content across the entire database is applied for this unit conversion. Fish centration of different fish species, PUE at the culture-system level, and PRE from name and auxiliary information such as living habitat, if not reported, are obtained controlled feeding experiments (see Methods: Data and Supplementary Discussion: from FishBase (http://www.fishbase.org). Supplementary Figs. 3 and 4 display the Data pattern for details). Supplementary Fig. 15 is a schematic diagram showing distributions of whole-body P concentrations. the global fishery P budget calculation. Harvested fish P mass is directly calculated PUE is defined as the proportion of supplementary P applied to feed fish and from fish live biomass and P concentration (per unit live biomass) for different fertilize the aquatic ecosystems that are recovered in harvested fish. Culture- species (Methods: P budget). Total aquaculture P-input can be estimated through system-level PUE quantifies the actual P investment considering various total aquaculture area, and feed and fertilizer inputs per unit area28, or through fish environmental and practical factors. The PUE database contains 168 cases from biomass production and the ratio of harvested to input P (phosphorus-use effi- 96 peer-reviewed publications that directly track culture-system-level P budget, ciency, PUE). We apply the latter approach due to the lack of a direct survey of including, at least, P harvested through fish and P input through feeds or/and global aquaculture area, and feed and fertilizer P input per unit area. Ref. 52 modeled finfish aquaculture P-input of major finfish species based on harvested fertilizers (Supplementary Data 2). Culture systems cover pond, tank, cages, recirculating, and flow-through aquaculture systems across India, United States, fish P mass, feed conversion ratio (FCR, the ratio of feed biomass to fish biomass), Madagascar, Thailand, Mexico, Brazil, Ireland, China, Honduras, Czech Republic, and P fraction in feed. Here we use a different approach to estimate aquaculture Sweden, Bangladesh, Israel, Australia, France, Vietnam, Poland, and Saudi Arabia. P-input based on harvested fish P mass and PUE at culture-system level (Methods: Feeding experiments frequently quantify the ratio of P that is incorporated into fish P budget). We do this, first because aquaculture is highly heterogeneous. Culture- biomass to the total fish P intake (PRE, Supplementary Data 3) at a single fish- system-level PUE takes into account diverse aquaculture practices in the real world, species level. We do not incorporate feeding experiments into our PUE database if which might deviate from PUE or P retention from idealized feeding experiments the study does not track the P budget for the whole culture system. Species-level focusing mostly on a single species (Figs S5 and S6). Second, the PUE at culture- system-level approach avoids the use of additional variables (such as P fertilizer PUE is close to culture-system-level PUE in monoculture. However, polyculture systems are a common practice where culture-system-level PUE might deviate input rate and P fraction in different feeds) for which insufficient data are available. from species-level PUEs. For example, four of the most widely raised species, silver We further separate the global budgets by continents and countries, and project carp (a photoplankton filter feeder), grass carp (a herbivorous macrophyte feeder), into the future. The PRE database is applied as a reference to assess whether a common carp (an omnivorous detritus feeder), and bighead carp (a zooplankton P-neutral fishery is technically achievable in the future. filter feeder) are frequently cultured together. Environmental conditions affect the portion of feed that is not taken up by fish, and waste from one species can serve as Data. World fishery production (1950–2016) is obtained by combining two food for another in polyculture. The P pre-existing in water used to raise fish is not databases, the Food and Agriculture Organization of the United Nations (FAO) considered as an external P input because we focus on land–aquatic P exchanges Global Fishery Production database, FishStatJ version 3.04.625 and the recon- (Supplementary Methods). Supplementary Fig. 5 shows the distribution of culture- structed wild marine fish capture database from Sea Around Us (http://www. system-level PUE. seaaroundus.org/). FishStatJ 3.04.6 tracks the annual nominal capture of wild finfish, crustaceans P budget. Annual P-harvest is estimated from Eq. (1) by summing P mass from and mollusks, the harvest of aquatic plants and mammals, as well as fish harvested each harvested species. from aquaculture and other fish farming from fresh, brackish, and marine waters. The database separates production from different countries, for multiple species, X n Phat ¼ Wi  R i ; ð1Þ and across diverse water areas around the globe from 1950 to 2016. In line with the i¼1 FAO annotation, fish in this study refers to finfishes, crustaceans, mollusks, and other aquatic invertebrates. Total fishery production increased from 21 Tg in 1950 where Phat (Tg P yr−1) is the annual P-harvest, Wi (Tg yr−1) is the weight of fish to 205 Tg (including aquatic plants and mammals) in 201625, while finfish, live biomass production, and Ri is whole-body P concentration (P mass fraction per crustaceans, and molluk production increased from 19 Tg in 1950 to 169 Tg in live biomass) for each species i. The mean whole-body P concentration is used for 201625. Aquatic plants, overwhelmingly seaweeds, can remove nutrients from the each fish species, averaging over a variety of living conditions, ontogenetic varia- aquatic ecosystems because farming of aquatic plants does not require large tions, and physiological status. For species that have no corresponding whole-body supplementary P input. However, with a relatively low P concentration, aquatic P concentration compiled in the whole-body P database, the mean P concentration NATURE COMMUNICATIONS | (2020)11:355 | https://doi.org/10.1038/s41467-019-14242-7 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-14242-7 from the same taxonomic order was substituted as a proxy. The whole-body data coverage, we cannot detect significant differences (p > 0.1, Student’s t test) in P database covers >80% of fish production at the taxonomic rank of order or lower. PUE between rearing systems (e.g., pond or cage), between countries, or between For the remaining 20%, the major group level (finfish, crustacean, and mollusk) freshwater versus marine aquaculture, with the exception that finfish farming has a mean whole-body P is applied. For example, cartilage fish (e.g., sharks, rays, and significantly (p < 0.01, Student’s t test) higher PUE compared with the farming of chimeras), which may be structurally different from bone fish, comprise around 1% crustacean species. Advancements in feeding technologies can improve feed of the total captured fish biomass; they are not included in our whole-body conversion ratio (FCR, the ratio of feed biomass to fish biomass) and thereby P database and the average finfish P concentration is used as an estimate. The class reduce feed requirement; however, in practice, its impact on PUE depends on “other aquatic invertebrates” covers fish that do not belong to any of the three economic, policy, and societal drivers such as the farmers’ socioeconomic status, major groups reported on by FAO; they account for less than 0.7% of total fish attitude, and behavior. Our PUE database does not support the occurrence of a biomass and are assumed to have a mean P concentration that is the same as that statistically significant shift in PUE through time. Because fish production of the entire database. databases report fish live weight separately for fresh and marine (including Uncertainty in harvested P estimation caused by variations in whole-body brackish water) environments, we also differentiate PUE between fresh and marine P concentration and fish biomass estimation is assessed by the Monte Carlo systems and among finfish, crustacean, and mollusk species. The culture-system- method. We randomly sample the whole-body P concentration 1000 times for each level PUE database is separated into six groups: freshwater finfish, marine water species in our database and calculate P-harvest budget. The number of compiled finfish, freshwater crustacean, marine water crustacean, freshwater mollusk, and whole-body P concentration entries differs for different species. P concentration for marine water mollusk. For each group j, total feed and fertilizer P input is each species may not cover the whole range of uncertainty. To avoid under- estimated by dividing harvested P mass (Phat,j) by culture-system-level PUE (PUEj, estimation of uncertainty associated with P concentration, we also make use of Eq. 2). Global feed and fertilizer P input is the sum of P inputs across these six information from higher taxonomic levels. If the number of P concentration entries groups. The distribution of PUE deviates from a normal distribution are no more than a cutoff criterion (one record as a start), then the random (Supplementary Fig. 5) and the median provides a better representation of typical samples for the corresponding species are generated from the selection pool with PUE. Currently, PUE of marine water mollusk is used for freshwater mollusk as we P concentration information covered by the taxonomic order that this species do not have experimental studies on PUE of this latter category in our database belongs to. P budgets are not sensitive to the cutoff criterion that determines the X 6 population from which random samples are drawn (see Supplementary Discussion: Pinp ¼ Phat; j =PUEj ; ð2Þ Uncertainty and sensitivity); we therefore report results from setting the cutoff j¼1 criterion to be one entry. When the species-level average is different to the average of the taxonomic order level, random samples taken directly from the bigger where Pinp is the P-input. Factors that may affect PUE were not studied taxonomic order pool would deviate the average P concentration from the species- individually but are taken into consideration in the uncertainty estimation. level pool. In order to keep the same average while covering a reasonable We conduct 1000 Monte Carlo calculations randomly sampling PUE (within uncertainty range, we conduct normal sampling using the species-level average and each group) through the compiled PUE database that incorporates a range of the standard deviation that is rescaled by the coefficient of variation at the variation (see also Supplementary Discussion: Uncertainty and sensitivity). taxonomic order level. Similarly, when the taxonomic order level P entry is no The final P budget uncertainties are characterized by the interquartile ranges more than the cutoff criterion, the scope of the random sampling lies in the major (IQR), because the distribution of PUE is highly skewed, and broader ranges are group. Here, we prefer to use mean, instead of median, to represent the average more likely to be affected by extreme values partly associated with the limited P concentration, as the median might be biased by the number of Monte Carlo sample size of our database. samples when the population size is small. The random sampling strategy is evaluated by changing the criterion based on which information is extracted from Reporting summary. Further information on research design is available in the order or major group level (Supplementary Discussion: Uncertainty and the Nature Research Reporting Summary linked to this article. sensitivity; Supplementary Fig. 9). Uncertainty or accuracy of fish biomass production data is debatable, and is not straightforward to assess75–79. One alternative strategy is to assign scores, Data availability which are associated with certain categorized uncertainty levels, for each data Databases of fish P concentration, culture-system-level P use efficiency, and P retention source that contributes to fish biomass production based on expert judgments78. efficiency are provided as Supplementary Data. Additional data that support the findings We do not trace back the uncertainty to each data source as it is hard to establish of this study are available from the corresponding author upon request to the a common standard for the scoring system from different countries and based on corresponding author. The source data underlying Figs. 2 and 3 are provided as a Source a variety of reporting sources. Instead, we look into the uncertainty by assuming Data file. a range of uncertainty levels (0%, 50%, and 100%) to assess the contribution of biomass uncertainty to the overall uncertainty estimation of P budgets (Supplementary Figs 9 and 10). A 50% percentage uncertainty corresponds to Code availability very low confidence, that is (quoting the IPCC80) “less than high agreement and Calculations were conducted through Python 2.7.15 and the code is available upon less than robust evidence”. As a conservative strategy, we report results at the request to the corresponding author. 50% uncertainty interval in the main text and the sensitivity of P budgets to fish biomass uncertainty is provided in Supplementary Discussion: Uncertainty and sensitivity. We randomly sample fish biomass within the corresponding Received: 6 July 2019; Accepted: 19 December 2019; percentile uncertainty and conduct 1000 Monte Carlo calculations for each uncertainty level. A wide range of feeds and fertilizers, varying in ingredients and weight, is applied in aquaculture81. Fish feed includes commercially manufactured compound aquafeeds with different additives, farm-made feeds from crop–livestock products and/or by-products, such as animal liver, blood, meat, References offal, poultry feathers, bone, and eggshell, but also barley, cereals, maize, rye, 1. Elser, J. J. Phosphorus: a limiting nutrient for humanity? Curr. Opin. sorghum, wheat, soybean, oilseed, coconut, cottonseed, and terrestrial invertebrates Biotechnol. 23, 833–838 (2012). (silkworm pupae, maggots, soldier flies, locusts, termites, and earthworms), and 2. Falkowski, P. et al. The global carbon cycle: a test of our knowledge of earth as fish with low economic value or by-products from aquaculture industry81. Species, a system. Science 290, 291–296 (2000). such as silver and bighead carps, filter-feeding fish species (e.g., bivalve mollusks) 3. Cordell, D. & White, S. Peak phosphorus: clarifying the key issues of a require no external feeds and are not fed. Instead, P fertilizers are applied to vigorous debate about long-term phosphorus security. Sustainability 3, increase aquatic primary productivity to support those species that do not rely on 2027–2049 (2011). external feeds. Fertilizers cover inorganic chemical fertilizers and manures from a 4. Obersteiner, M., Peñuelas, J., Ciais, P., van der Velde, M. & Janssens, I. A. The variety of sources, such as poultry, ruminant waste, swine waste, and human phosphorus trilemma. Nat. Geosci. 6, 897–898 (2013). excreta81. 5. Bouwman, A. F. et al. Lessons from temporal and spatial patterns in global use Drawing up a direct bottom-up estimate of feed and fertilizer P investment is of N and P fertilizer on cropland. Sci. Rep. 7, https://doi.org/10.1038/ difficult due to scarce data. Globally, total commercially manufactured compound srep40366 (2017). aquafeed is estimated to be 34.4–39.6 Tg in 201248. However, the global estimates of feed from farm-made products and low-value fish are largely undocumented. 6. Lun, F. et al. Global and regional phosphorus budgets in agricultural systems Farm-made aquafeeds are roughly reported to be between 18.7 and 30.7 Tg in and their implications for phosphorus-use efficiency. Earth Syst. Sci. Data 10, 200682, and low-value fish is estimated to be 5.6–8.8 Tg. Insufficient data are 1–18 (2018). available to quantify the global total of mineral fertilizer applied to aquaculture 7. Peñuelas, J. et al. Human-induced nitrogen-phosphorus imbalances alter globally. The volume of manure that goes into aquaculture and its phosphorus natural and managed ecosystems across the globe. Nat. Commun. 4, https:// concentration are also largely unrecorded. doi.org/10.1038/ncomms3934 (2013). Therefore, we estimate external aquaculture P input through culture-system- 8. Correll, D. L. The role of phosphorus in the eutrophication of receiving waters: level PUE (per feed and fertilizer applied). Due to the large variation and limited a review. J. Environ. Qual. 27, 261–266 (1998). 8 NATURE COMMUNICATIONS | (2020)11:355 | https://doi.org/10.1038/s41467-019-14242-7 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-14242-7 ARTICLE 9. Schindler, D. W. et al. Eutrophication of lakes cannot be controlled by the 20th century despite increased retention along the aquatic continuum. reducing nitrogen input: Results of a 37-year whole-ecosystem experiment. Biogeosciences 13, 2441–2451 (2016). Proc. Natl Acad. Sci. USA 105, 11254–11258 (2008). 39. Seitzinger, S. P. et al. Global river nutrient export: a scenario analysis of past 10. Smith, V. H., Tilman, G. D. & Nekola, J. C. Eutrophication: impacts of excess and future trends. Global Biogeochem. Cycles 24, https://doi.org/10.1029/ nutrient inputs on freshwater, marine, and terrestrial ecosystems. Environ. 2009gb003587 (2010). Pollut. 100, 179–196 (1999). 40. Bennett, E. M., Carpenter, S. R. & Caraco, N. F. Human impact on erodable 11. Sardans, J., Rivas-Ubach, A. & Peñuelas, J. The elemental stoichiometry of phosphorus and eutrophication: A global perspective. Bioscience 51, 227–234 aquatic and terrestrial ecosystems and its relationships with organismic (2001). lifestyle and ecosystem structure and function: a review and perspectives. 41. Galbraith, E. D., Carozza, D. A. & Bianchi, D. A coupled human-Earth model Biogeochemistry 111, 1–39 (2012). perspective on long-term trends in the global marine fishery. Nat. Commun. 8, 12. Carpenter, S. R. & Bennett, E. M. Reconsideration of the planetary boundary https://doi.org/10.1038/ncomms14884 (2017). for phosphorus. Environ. Res. Lett. 6, https://doi.org/10.1088/1748-9326/6/1/ 42. Guo, K., Zhao, W., Jiang, Z. Q. & Dong, S. L. A study of organic carbon, 014009 (2011). nitrogen and phosphorus budget in jellyfish-shellfish-fish-prawn polyculture 13. Rockstrom, J. et al. A safe operating space for humanity. Nature 461, 472–475 ponds. Aquac. Res. 48, 68–76 (2017). (2009). 43. Vass, K. K., Wangeneo, A., Samanta, S., Adhikari, S. & Muralidhar, M. 14. O’Neill, D. W., Fanning, A. L., Lamb, W. F. & Steinberger, J. K. A good life for Phosphorus dynamics, eutrophication and fisheries in the aquatic ecosystems all within planetary boundaries. Nat. Sustainability 1, 88–95 (2018). in India. Curr. Sci. 108, 1306–1314 (2015). 15. Ringeval, B. et al. Phosphorus in agricultural soils: drivers of its distribution at 44. David, F. S., Proenca, D. C. & Valenti, W. C. Phosphorus budget in integrated the global scale. Glob. Change Biol. 23, 3418–3432 (2017). multitrophic aquaculture systems with Nile Tilapia, Oreochromis niloticus, 16. Roy, E. D. et al. The phosphorus cost of agricultural intensification in the and Amazon River Prawn, Macrobrachium amazonicum. J. World Aquac. Soc. tropics. Nat. Plants 2, https://doi.org/10.1038/nplants.2016.43 (2016). 48, 402–414 (2017). 17. Bouwman, A. F., Beusen, A. H. W. & Billen, G. Human alteration of the global 45. Bostock, J. et al. Aquaculture: global status and trends. Philos. Trans. R. Soc. B: nitrogen and phosphorus soil balances for the period 1970–2050. Global Biol. Sci. 365, 2897–2912 (2010). Biogeochem. Cycles 23, https://doi.org/10.1029/2009gb003576 (2009). 46. Bouwman, L. et al. Exploring global changes in nitrogen and phosphorus 18. Steinfeld, H. et al. Livestock’s Long Shadow: Environmental Issues and Options cycles in agriculture induced by livestock production over the 1900-2050 (United Nations Food and Agriculture Organization, 2006). period. Proc. Natl Acad. Sci. USA 110, 20882–20887 (2013). 19. Chowdhury, R. B., Moore, G. A., Weatherley, A. J. & Arora, M. Key sustainability 47. Chen, M. P. & Graedel, T. E. A half-century of global phosphorus flows, challenges for the global phosphorus resource, their implications for global food stocks, production, consumption, recycling, and environmental impacts. Glob. security, and options for mitigation. J. Clean. Prod. 140, 945–963 (2017). Environ. Change-Hum. Policy Dimens. 36, 139–152 (2016). 20. Schoumans, O. F. et al. Mitigation options to reduce phosphorus losses from 48. Tacon, A. G. J. & Metian, M. Feed matters: satisfying the feed demand of the agricultural sector and improve surface water quality: a review. Sci. Total aquaculture. Rev. Fish. Sci. Aquac. 23, 1–10 (2015). Environ. 468, 1255–1266 (2014). 49. Islam, M. S. Nitrogen and phosphorus budget in coastal and marine cage 21. Bene, C. et al. Contribution of fisheries and aquaculture to food security and aquaculture and impacts of effluent loading on ecosystem: review and analysis poverty reduction: assessing the current evidence. World Dev. 79, 177–196 (2016). towards model development. Mar. Pollut. Bull. 50, 48–61 (2005). 22. Bene, C. et al. Feeding 9 billion by 2050-Putting fish back on the menu. Food 50. Sharpley, A. et al. Phosphorus legacy: overcoming the effects of past Security 7, 261–274 (2015). management practices to mitigate future water quality impairment. J. Environ. 23. Tidwell, J. H. & Allan, G. L. Fish as food: aquaculture’s contribution— Qual. 42, 1308–1326 (2013). ecological and economic impacts and contributions of fish farming and 51. Bouwman, A. F. et al. Global hindcasts and future projections of coastal capture fisheries. EMBO Rep. 2, 958–963 (2001). nitrogen and phosphorus loads due to shellfish and seaweed aquaculture. Rev. 24. Maranger, R., Caraco, N., Duhamel, J. & Amyot, M. Nitrogen transfer from Fish. Sci. 19, 331–357 (2011). sea to land via commercial fisheries. Nat. Geosci. 1, 111–113 (2008). 52. Bouwman, A. F. et al. Hindcasts and future projections of global inland and 25. F. A. O. Fishery and Aquaculture Statistics. Global production by production coastal nitrogen and phosphorus loads due to Finfish aquaculture. Rev. Fish. source 1950–2016. In FAO Fisheries and Aquaculture Department [online]. Sci. 21, 112–156 (2013). Rome. Updated 2018. www.fao.org/fishery/statistics/software/fishstatj/en (2018). 53. Edwards, P. Aquaculture environment interactions: Past, present and likely 26. Waite, R. et al. Improving productivity and environmental performance of future trends. Aquaculture 447, 2–14 (2015). aquaculture World Resource Institute (ed.) (Creating a Sustainable Food 54. Costa-Pierce, B. Ecological Aquaculture: The Evolution of the Blue Revolution Future, Washington DC, 2014). (Wiley, 2002). 27. Froehlich, H. E., Runge, C. A., Gentry, R. R., Gaines, S. D. & Halpern, B. S. 55. Huang, C. L. et al. Changing phosphorus metabolism of a global aquaculture Comparative terrestrial feed and land use of an aquaculture-dominant world. city. J. Clean. Prod. 225, 1118–1133 (2019). Proc. Natl Acad. Sci. USA 115, 5295–5300 (2018). 56. Zou, L. & Huang, S. Chinese aquaculture in light of green growth. Aquac. Rep. 28. Hall, S. J., Delaporte, A., Phillips, M. J., Beveridge, M. & O’Keefe, M. Blue 2, 46–49 (2015). Frontiers: Managing the Environmental Costs of Aquaculture (The WorldFish 57. Wang, Q. D. et al. Paradigm changes in freshwater aquaculture practices in Center, Penang, Malaysia, 2011). China: moving towards achieving environmental integrity and sustainability. 29. Naylor, R. L. et al. Feeding aquaculture in an era of finite resources. Proc. Natl Ambio 47, 410–426 (2018). Acad. Sci. USA 106, 15103–15110 (2009). 58. Cordell, D., Rosemarin, A., Schroder, J. J. & Smit, A. L. Towards global 30. Troell, M. et al. Does aquaculture add resilience to the global food system? phosphorus security: a systems framework for phosphorus recovery and reuse Proc. Natl Acad. Sci. USA 111, 13257–13263 (2014). options. Chemosphere 84, 747–758 (2011). 31. Verdegem, M. C. J. Nutrient discharge from aquaculture operations in 59. Arvanitoyannis, I. S. & Kassaveti, A. Fish industry waste: treatments, function of system design and production environment. Rev. Aquac. 5, environmental impacts, current and potential uses. Int. J. Food Sci. Technol. 158–171 (2013). 43, 726–745 (2008). 32. Zhang, Y., Bleeker, A. & Liu, J. G. Nutrient discharge from China’s 60. Cao, L. et al. China’s aquaculture and the world’s wild fisheries. Science 347, aquaculture industry and associated environmental impacts. Environ. Res. 133–135 (2015). Lett. 10, https://doi.org/10.1088/1748-9326/10/4/045002 (2015). 61. Mo, W. Y., Man, Y. B. & Wong, M. H. Use of food waste, fish waste and food 33. FAO. The State of World Fisheries and Aquaculture. Contributing to Food Security processing waste for China’s aquaculture industry: needs and challenge. Sci. and Nutrition for All (The State of World Fisheries and Aquaculture, 2016). Total Environ. 613, 635–643 (2018). 34. Doughty, C. E. et al. Global nutrient transport in a world of giants. Proc. Natl 62. Ghaly A. E., et al. Fish processing wastes as a potential source of proteins, Acad. Sci. USA 113, 868–873 (2016). amino acids and oils: a critical review. Microbial Biochem. Technol. 5, 107–129 35. Otero, X. L., De La Pena-Lastra, S., Perez-Alberti, A., Ferreira, T. O. & Huerta- (2013). Diaz, M. A. Seabird colonies as important global drivers in the nitrogen and 63. Tacon, A. G. J. & Metian, M. Global overview on the use of fish meal and fish phosphorus cycles. Nat. Commun. 9, https://doi.org/10.1038/s41467-017- oil in industrially compounded aquafeeds: trends and future prospects. 02446-8 (2018). Aquaculture 285, 146–158 (2008). 36. Wang, R. et al. Significant contribution of combustion-related emiscions to the 64. Ganrot, Z., Broberg, J. & Byden, S. International Conference on Nutrient atmospheric phosphorus budget. Nat. Geosci. 8, 48–54 (2015). Recovery from Wastewater Streams Vancouver (ed. Ashley, K, Mavinic, D., & 37. Mekonnen, M. M. & Hoekstra, A. Y. Global anthropogenic phosphorus loads Koch, F.) (IWA Publishing, London). to freshwater and associated grey water footprints and water pollution levels: a 65. Iatrou, E. I., Stasinakis, A. S. & Aloupi, M. Cultivating duckweed Lemna minor high-resolution global study. Water Resour. Res. 54, 345–358 (2018). in urine and treated domestic wastewater for simultaneous biomass 38. Beusen, A. H. W., Bouwman, A. F., Van Beek, L. P. H., Mogollon, J. M. & production and removal of nutrients and antimicrobials. Ecol. Eng. 84, Middelburg, J. J. Global riverine N and P transport to ocean increased during 632–639 (2015). NATURE COMMUNICATIONS | (2020)11:355 | https://doi.org/10.1038/s41467-019-14242-7 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-14242-7 66. Jimenez, J., Bott, C., Love, N. & Bratby, J. Source separation of urine as an Acknowledgements alternative solution to nutrient management in biological nutrient removal This work was financially supported by the IMBALANCE-P project of the European treatment plants. Water Environ. Res. 87, 2120–2129 (2015). Research Council (ERC-654 2013-SyG-610028). 67. Mihelcic, J. R., Fry, L. M. & Shaw, R. Global potential of phosphorus recovery from human urine and feces. Chemosphere 84, 832–839 (2011). 68. Amann, A. et al. Environmental impacts of phosphorus recovery from Author contributions Y.H., P.C., D.G., J.S. and J.P. designed this study. Y.H. and J.S. collected the data. Y.H. municipal wastewater. Resour. Conserv. Recycling 130, 127–139 (2018). conducted the analysis and drafted the paper. Y.H., P.C., D.G., J.S., J.P., F.C. and H.Z. 69. Egle, L., Rechberger, H., Krampe, J. & Zessner, M. Phosphorus recovery from discussed the results and contributed to the paper. municipal wastewater: an integrated comparative technological, environmental and economic assessment of P recovery technologies. Sci. Total Environ. 571, 522–542 (2016). Competing interests 70. Baker, L. A. Can urban P conservation help to prevent the brown devolution? The authors declare no competing interests. Chemosphere 84, 779–784 (2011). 71. Metson, G. S. et al. Phosphorus in Phoenix: a budget and spatial representation of phosphorus in an urban ecosystem. Ecol. Appl. 22, 705–721 (2012). Additional information 72. Hamilton, H. A. et al. Investigating cross-sectoral synergies through integrated Supplementary information is available for this paper at https://doi.org/10.1038/s41467- aquaculture, fisheries, and agriculture phosphorus assessments: a case study of 019-14242-7. Norway. J. Ind. Ecol. 20, 867–881 (2016). 73. Fry, J. P. et al. Environmental health impacts of feeding crops to farmed fish. Correspondence and requests for materials should be addressed to Y.H. Environ. Int. 91, 201–214 (2016). 74. Blanchard, J. L. et al. Linked sustainability challenges and trade-offs among Peer review information Nature Communications thanks Christopher Anderson, fisheries, aquaculture and agriculture. Nature Ecology &. Nat. Ecol. Evolution Michael Vanni and the other, anonymous, reviewer(s) for their contribution to the peer 1, 1240–1249 (2017). review of this work. Peer reviewer reports are available. 75. Pauly, D. & Zeller, D. Comments on FAOs State of World Fisheries and Aquaculture (SOFIA 2016). Mar. Policy 77, 176–181 (2017). Reprints and permission information is available at http://www.nature.com/reprints 76. Pauly, D. & Zeller, D. The best catch data that can possibly be? Rejoinder to Ye et al. “FAO’s statistic data and sustainability of fisheries and aquaculture”. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in Mar. Policy 81, 406–410 (2017). published maps and institutional affiliations. 77. Ye, Y. M. et al. FAO’s statistic data and sustainability of fisheries and aquaculture: comments on Pauly and Zeller (2017). Mar. Policy 81, 401–405 (2017). 78. Pauly, D. & Zeller, D. Catch reconstructions reveal that global marine fisheries Open Access This article is licensed under a Creative Commons catches are higher than reported and declining. Nat. Commun. 7, https://doi. Attribution 4.0 International License, which permits use, sharing, org/10.1038/ncomms10244 (2016). adaptation, distribution and reproduction in any medium or format, as long as you give 79. Campbell, B. & Pauly, D. Mariculture: a global analysis of production trends appropriate credit to the original author(s) and the source, provide a link to the Creative since 1950. Mar. Policy 39, 94–100 (2013). Commons license, and indicate if changes were made. The images or other third party 80. IPCC. Climate Change 2007: The Physical Science Basis. Contribution of material in this article are included in the article’s Creative Commons license, unless Working Group I to the Fourth Assessment Report of the Intergovernmental indicated otherwise in a credit line to the material. If material is not included in the Panel on Climate Change (eds Solomon, S., et al.) 996 (Cambridge University article’s Creative Commons license and your intended use is not permitted by statutory Press, Cambridge, UK and New York, NY, USA, 2007). regulation or exceeds the permitted use, you will need to obtain permission directly from 81. Tacon, A. G. J., Metian, M. & Hasan, M. R. Feed ingredients and fertilizers for the copyright holder. To view a copy of this license, visit http://creativecommons.org/ farmed aquatic animals: sources and composition. FAO Fisheries and Aquaculture licenses/by/4.0/. Technical Paper. No. 540. 209 (FAO Fisheries and Aquaculture, Rome, 2009). 82. Tacon, A. G. J. H. M. R. et al. In Proc. Global Conference on Aquaculture 2010. Farming the Waters for People and Food (eds Subasinghe, R. P., et al.) (FAO, Rome, Italy 2013). © The Author(s) 2020 10 NATURE COMMUNICATIONS | (2020)11:355 | https://doi.org/10.1038/s41467-019-14242-7 | www.nature.com/naturecommunications