Feed-food competition in global aquaculture: Current trends and prospects

Feed-food competition is the allocation of resources that can be used to feed humans to animal feed instead, a current but unsustainable practise not well documented for aquaculture. Here, we analysed feed-food competition in aquaculture using two measures; natural trophic levels (TLs) and species-specific human-edible protein conversion ratios (HePCRs). The HePCR equals the ratio of human edible protein in feed (input) to the human edible protein in animal produce (output). To provide prospects on aquaculture's potential to convert human inedible by-products into edible biomass, data on aquaculture production were collected and categorized based on natural TLs. HePCRs were computed for four aquaculture species produced in intensive aquaculture systems: Atlantic salmon, common carp, Nile tilapia and whiteleg shrimp. Under current feed use, we estimated that the carp, tilapia and shrimp considered were net contributors of protein by requiring (cid:1) 0.6 kg of human edible protein to produce 1 kg of protein in the fillet/meat. Considering soya bean meal and fish-meal as food-competing ingredients increased the HePCR to (cid:1) 2 and turned all of the case-study species into net consumers of protein. To prevent this increase, the use of high-quality food-competing ingredients such as fishmeal, or soya bean products should be minimized in aquaculture feed. In the future, the role of aquaculture in circular food systems will most likely consist of a balanced mix of species at different TLs and from different aquaculture systems, depending on the by-products available.


| INTRODUCTION
Humanity is facing the challenge of feeding the world's growing population while staying within planetary environmental boundaries.
As the current food system exceeds many planetary boundaries, a promising way forward is to build circular food systems. In which biomass from arable land and water bodies is prioritized for human food and other basic needs rather than for animal feed, thus reducing feed-food competition. [1][2][3] In this paradigm, farm animals, including aquaculture species, should not consume human edible biomass but instead convert by-products from crops, livestock and fisheries that are inedible for humans, into edible biomass. In addition to these by-products, animals in circular food systems can also convert plantbased food waste and grass resources into food. Geert F. Wiegertjes and Imke J. M. de Boer shared equally to the last authorship.
To date, discussions of feed-food competition in aquaculture have largely addressed the use of fishmeal, and for a good reason: 9% of fisheries landings are transformed into fishmeal and fish oil, 4 of which 90% can be considered food-grade. 5 These studies often focused on the fish-in: fish-out ratio (FIFO) as an indicator for feed efficiency of aquaculture species. The FIFO differs greatly among species. For example, in 2009, salmon had a FIFO of around 5 and shrimp around 1.5. 6 Over the years, the FIFO has decreased to around 0.3 for global aquaculture. 7 An important reason for this decrease is that there is a growing trend to replace fishmeal with plant-based protein sources, such as soy protein concentrate. 8,9 These replacement ingredients, however, also often cause feed-food competition because they can be used directly as food. 8,9 In addition, plant-based ingredients, such as soya-bean-based ingredients, could influence feed-food competition indirectly by increasing land use for the production of animal feed instead of human food.
Nevertheless, fish and shellfish are rich in macronutrients (e.g., protein, fat) and micronutrients (e.g., calcium, iron, zinc, selenium; vitamins A, B and D; iron) [10][11][12] and can be a valuable addition to a healthy diet even when consumed in small amounts. 10 In fact, fish is the main source of essential omega-3 long-chain polyunsaturated fatty acids in human diets. 13 Furthermore, farmed fish convert feed into food relatively efficiently. 14 For example, Atlantic salmon has a feed conversion efficiency similar to that of chicken, and both are better than those of most other farmed animals. 15 To determine the unique role of aquaculture in the transition towards healthy and circular food systems, more insight is needed into feed-food competition in global aquaculture.
One way to do so is to explore an animal's natural ability to upgrade specific by-products into food. This ability is determined by animal species, breed, and production-system intensity, and varies greatly among the more than 400 fish and shellfish species farmed in aquaculture. Here, we used the natural trophic level (TL) as an indicator of an animals' natural ability to upgrade specific by-products from human food. 9,15 Aquatic species at low TLs, such as primary producers (algae) and filter feeders (e.g., mussels, oysters) do not consume human edible biomass and therefore do not contribute to feed-food competition. Similarly, herbivorous species at low TLs have enzymes that can digest dietary fibres and other carbohydrates that humans cannot. Aquaculture species at higher TLs-omnivores and carnivores species-have diets that are more similar to those of humans and are well adapted to convert fish or other animal-based by-products into food. We therefore hypothesize that feed-food competition increases as the natural TL in aquaculture increases. To date, however, insight into the ability of individual aquaculture species from different natural TLs to convert by-products into food is lacking.
Another way to gain insight into feed-food competition is to use the HePCR to quantify the net contribution of farmed fish to the supply of human edible protein. The HePCR equals the ratio of human edible protein in feed (input) to the human edible protein in the animal product (output). Much quantification of HePCR has focused on livestock. In general, monogastric animals consume more protein than they produce (i.e., HePCR > 1) while grass-fed ruminants can produce more protein than they consume (i.e., HePCR < 1). HePCRs are reported for a variety of animals in several countries, 16 including England (beef, lamb, pig, poultry 17 ), Austria (beef, veal, swine, chicken, turkey, sheep, goat 18 ) and Ireland (beef, pig 19 ), but have not been estimated for aquaculture species. In aquaculture, only the human edibility of the feed has been estimated. Insights into the conversion efficiency of human edible feed protein into aquatic protein (i.e., quantifications of HePCR), however, are lacking.
Here, we analysed feed-food competition in aquaculture using both criteria: natural TLs and species-specific HePCR. We addressed the current status and trends in aquaculture production based on TLs and calculated HePCRs of current intensive aquaculture systems. The HePCRs were calculated for case studies of four key aquaculture species: Atlantic salmon, common carp (Cyprinus carpio), whiteleg shrimp (Litopenaeus vannamei) and Nile tilapia (Oreochromis niloticus).

| Global aquaculture production
To contextualize our findings about feed-food competition, we first reviewed generic data on aquaculture production by retrieving production (wet weight) and economic data (USD) at global and continental levels from FishstatJ. 20 To develop an overview of current aquaculture production (reference year 2019), we selected the 50 species produced most (in wet weight). 20

| Natural trophic levels
We first quantified the relative contribution of each TL to current global aquaculture production (in kg wet weight and edible protein), aggregated by species groups. To this end, we categorized production data of the 50 aquaculture species produced most by TL range. Each species' natural TL (1-5) was based on information from FishBase 21 ( Figure 1), shown to be a good data source for the trophic ecology of finfish. 22 Because FishBase does not include data on molluscs or crustaceans, their natural TLs were extracted from the literature (refer to Tables A1 and A2 for details). If the TL of a species was unknown, that of a closely related species was assumed to apply. As primary producers, all seaweeds were assigned to TL 1. We expressed the relative contribution of each TL range to current global aquaculture in terms of edible protein by multiplying each species' production volume (wet weight) by its edible yield and protein content. Edible yield was defined as fillet yield for finfish species and as meat yield for molluscs and crustaceans, as shells and exoskeletons van RIEL ET AL.
were considered inedible. Because seaweeds have other uses besides food, we multiplied total seaweed production per species by the proportion used as human food, which was derived from Naylor et al. 24 Seaweed produced for food was assumed to be completely human edible. If the edible yield of a species was unknown, that of a closely related species was assumed (refer to Tables A1 and A2 for details).
The protein content of individual fish species was collected from the U.S. Department of Agriculture (USDA) database. 25 When data were not available from the USDA database, 25 they were obtained from FishBase, 21 standard tables of food composition in Japan 26 or the literature (see Tables A1 and A2 for details). Protein contribution was based on raw fish, and cooking losses were not included.
We then quantified the trend in the mean TL of aquaculture production globally and by continent from 1980-2019. We collected data on annual production and the natural TL of the 25 species produced most (wet weight) globally and categorized them by continent.
The method for collecting data on production, natural TL, and the grouping of species was the same as that used for analysing the global natural TL.

| Human edible protein conversion ratios
To embed HePCR results in the existing aquaculture literature, we also calculated the protein conversion ratio (PCR), as an indicator of feed efficiency.

| Case studies
To obtain an initial impression of feed-food competition in aquaculture at the species level, calculations were performed for four casestudy species-three finfish (one from each natural TL) and one crustacean-chosen based on two criteria: having the highest economic value and being produced in intensive systems. Consequently, we selected: Atlantic salmon (TL 4-5), common carp (TL 3-4), Nile tilapia (TL 2-3) and whiteleg shrimp (TL 2-3).

| Data sources
PCR, HePCR e and HePCR d were calculated for each species (Box 1).
For the ingredients in the diets, we focused on the grow-out phase.
Diets for common carp, Nile tilapia and whiteleg shrimp were based on confidential surveys with five people active in the aquaculture feed industry. They were asked to estimate the feed composition for each species for intensive cultivation in 2020. The mean of the suggested diets was used as diet composition in this study ( Table 1). The diet composition of Atlantic salmon was based on the feed used for Atlantic BOX 1 Protein conversion ratio (PCR), human edible protein conversion ratio (HePCR e ) and human digestible protein conversion ratio (HePCR d ) adapted from Laisse et al. 27  where i the aquaculture species considered, j the ingredients of the diets consumed by the animal. FCR is the feed conversion ratio of the animal, Feed the proportion of the feed that is ingredient ( j), CP is the crude protein content (%), HE the human edibility and DIAAS the digestible indispensable amino acid score (%).
salmon aquaculture in Norway in 2020. 28 The assumed range in feed conversion ratio (FCR) of each species was based on FCRs found for these species for intensive production systems in the literature (  9 and Mottet et al. 16 ( Table 2). Both studies reported the same human edibility for most ingredients, except for soya bean meal and fishmeal. Sandström et al. 9 considered fishmeal made from whole fish food-competing but fishmeal made from fish by-products non-food-competing, while Mottet et al. 16 considered all fishmeal non-food-competing. Furthermore, while soy protein concentrate is food-competing, soya bean meal is considered non-food-competing because it is a by-product of soya bean oil production, and the meal is used almost entirely as a feed ingredient. However, soya bean differs from all other ingredients in that its by-product (soya bean meal) is the primary driver of soya bean production. As a result, soya bean meal can be considered an indirect competitor to human food, 16 which Mottet et al. 16 did, but Sandstrom et al. 9 did not. To capture these differences, we created two scenarios, in which soya bean meal and fishmeal were considered (1) non-food-competing or (2) food-competing. For fishmeal, we used a human edibility of 0.66 instead of 1 to represent that 27% of fishmeal is made up of by-products (non-food) 4 and that 90% of the whole fish used to create fishmeal can be considered food grade. 5 For the human edibility of the output of the species, we focused on the current habits of eating primarily fillets. To represent protein quality, we included the digestible indispensable amino acid score (DIAAS) for the selected fish Animal-based ingredients Distillers grains - species and the feed ingredients ( Table 2). The Food and Agriculture Organization of the United Nations (FAO) recommends the DIAAS 41 as a measure of protein quality. The DIAAS reflects the content of the first limiting indispensable amino acid in a feed/food ingredient relative to the requirement for the same amino acid by humans. 41 We used the amino acid requirement pattern for a 6-month to 3-year-old child as the reference protein's amino acid profile, similar to studies by Laisse et al. 42 and Ertl et al., 39 and as recommended by the FAO. 41 The DIAASs of feed ingredients were extracted from Ertl et al. 39 or, if not included by them, estimated using their method. 37 Due to a lack of data on the human ileal amino acid digestibility of fish fillets, the DIAASs of the case-study species were calculated based on amino acid scores from the USDA, 25 assuming an amino acid digestibility of 94% for fillets/meat. 41 3 | RESULTS

| Trophic levels
The contribution of each aquaculture species groups to global aquaculture production in 2019 varied ( Figure 2).
In 2019, TL 2-3 produced the most wet weight (59%) and edible protein (60%). Although dominated by freshwater fish and molluscs, it was the TL range with the greatest diversity of species groups produced. In general, aquaculture species at lower TL (TL 1-3) contributed less to global protein production than to global aquaculture wet-weight volumes. For aquatic plants, this is due to their high water and low protein content whereas fish from low TLs have lower edible yields than at higher TLs.
Globally, over the past 40 years, the mean natural TL of aquaculture species increased slightly (Figure 3a), but it differed greatly among regions, especially China and Europe. In Europe, it increased due to the large increase in production of diadromous fish (especially Atlantic salmon) over the past 35 years. Since 2002, Europe was also the continent with the highest mean TL (Figure 3a,b).
In contrast, the mean TL of aquaculture species in Asia/China remained relatively low and stable over the past 30 years (1985-2015) due to higher growth of production of freshwater fish, molluscs and aquatic plants than that of other species groups (Figure 3c).

| Human edibility of fish diets
In scenario 1 (soya bean meal and fishmeal non-competing), Atlantic salmon had the highest percentage of food-competing ingredients in the diet (45%) (Figure 4a) because soy protein concentrate was the primary protein source. tilapia had a lower percentage of food-competing ingredients in their diets (3%, 12% and 12%, respectively), with wheat products providing most edible protein (Figure 4a).
In scenario 2 (soya bean meal and fishmeal food-competing), the percentage of human edible protein increased for all diets, ranging from 49% to 65% (Figure 4b), because soya bean meal was the main protein source in the diets of common carp, whiteleg shrimp and Nile tilapia and fishmeal was included in the diets of Atlantic salmon and whiteleg shrimp. Atlantic salmon and whiteleg shrimp had the highest percentages of food-competing ingredients in their diets (59% and 65%, respectively). Ingredients included in the aquaculture diets in relatively large percentages that were not food-competing included livestock by-products, gluten meals and cereal bran.

| Conversion ratios
The PCR ranged from 3.4 to 8.7 for the case-study species (Table 3).
Atlantic salmon converted protein the most efficiently, followed by whiteleg shrimp and then common carp and Nile tilapia, both of which had similar PCRs. Differences in PCR among species were caused by differences in the FCR, edible yield and protein content of the feed.
Compared to the other species, Atlantic Salmon had the lowest FCR and highest fillet/meat yield.

In scenario 1, Atlantic salmon was the only species with an
HePCR e greater than 1 (range: 1.5-2.0; Figure 5a) and thus consumed more human edible protein than it produced. The HePCR e of the other species lay below 1 (range: 0.2-1.0), implying that they produced more human-edible protein than they consumed. Values of HePCR d , which considers protein digestibility, were lower than those of HePCR e due to the relatively higher quality of fillet protein than feed protein ( Figure 5).
In scenario 2, all four species were net consumers of protein:

| DISCUSSION
Given the diversity of aquaculture species, the present study was intended as a starting point for exploring and analysing feed-food competition for additional species, systems and locations.

| Synthesis
We worked from the assumption that aquaculture offers great potential to produce food while avoiding feed-food competition. We used natural TLs as a starting point to analyse feed-food competition, as the natural ability of an animal to upgrade specific by-products into food can determine its role in a circular food system. In both Europe and the Americas, Atlantic salmon was the species at a high TL (TL 4-5), whose production was largest and grew the most rapidly, which drove the increase in average produced natural TL. Feeding compound feeds has generally resulted in aquaculture diets with an effective TL lower than that of natural diets (natural TL). 43 This decrease may appear positive if assuming that diets at lower TLs generally cause less feed-food competition because they include more plant-based ingredients and less fishmeal. However, when fishmeal is replaced by soy protein concentrate, as for salmon, the positive impact on feed-food competition is not apparent because soy protein concentrate is human edible and has higher protein quality (i.e., a higher DIAAS) than fishmeal. As a result, species at a naturally high TL, such as salmon, continue to receive relatively higher quality (plantbased) ingredients, resulting in highly human edible diets.
When investigating feed-food competition in the present study, classifying soya bean meal and fishmeal as either foodcompeting or non-food-competing ingredients had large influence on the net contribution to protein supply of the four aquaculture species. When we considered them as human-edible, not only did species at a high trophic (i.e., salmon) appear as net consumers of protein, but so did species at a lower TL, (i.e., common carp, whiteleg shrimp and Nile tilapia). Soya bean meal is the ingredient used most in aquaculture compound feeds. 8 Although soya bean meal itself is considered inedible, 9,16 its production is the main driver of land use. 16 Soya bean meal causes indirect feed-food competition, as the land used to produce soya bean meal could have been used to grow food crops for direct human consumption. 44 Although fishmeal does not require land its production can lead to overfishing and of greenhouse gas emissions. 45 Replacing soya bean meal in aquaculture feeds would reduce feed-food competition in aquaculture drastically.
When soya bean meal and fishmeal were considered foodcompeting, Nile tilapia, which has a low TL, had the highest HePCR, while Atlantic salmon, which has a high TL, had the lowest HePCR. This may seem surprising, because Atlantic salmon and whiteleg shrimp had the highest percentage of human-edible protein in their feed. However, the low HePCR of Atlantic salmon can be explained by its relatively high growth rate and feed efficiency, Human-edible protein ingredients as percentage of the total protein of the diets for the selected species. (a) Scenario 1 (fish meal and soya bean meal are non-food-competing ingredients) and (b) Scenario 2 (fish meal and soya bean meal are food-competing ingredients).
T A B L E 3 Feed conversion ratios (FCR), fillet/meat yield, protein contents and crude protein conversion ratio (PCR) for the four case-study species from literature and the present study. due, among other things, to years of selective breeding. 46 In addition, when kept in intensive systems, species at a naturally low TL such as Nile tilapia are often fed high-quality protein (e.g., soya bean meal) to increase growth rates and decrease FCR. Thus, intensive aquaculture systems do not optimally align with the natural ability of species at a low TLs to upgrade lower quality by-products or natural biomass. For these species, extensive systems as well as ecological intensification, for example, nutritious ponds, are better suited.

| Feed-food competition in aquaculture compared to livestock
Compared to livestock, absolute feed-food competition in aquaculture is relatively small, because aquaculture represents for only a small percentage ($1.2%) of global feed consumption, compared to that of cattle (73%), pigs (20%) and poultry (7%). 9 Looking only at global human-edible feed consumption, however, aquaculture represents a larger percentage (3.8%), 9 likely related to the relatively high protein requirements of fed aquaculture species.
Overall, most gain to reduce feed-food competition is to be made with livestock.
Directly comparing the HePCR of livestock and aquaculture is hampered by differences in metabolism and housing. The most logical comparison with monogastric species, such as poultry and pigs. When their HePCR ratios are compared, broilers (HePCR e $ 5.2) and industrially produced pigs (HePCR e $ 4.5) have higher HePCRs than the aquaculture species examined (HePCR e : 0.2-2). 16

| Limitations of this study
We attempted to approach feed-food competition scenarios in aquaculture using all available information and objective criteria, but improvements are always possible. First, information on complete feed formulation in this study was obtained for three of the four selected species by confidential surveys of people in the aquaculture feed industry. Feed formulation, however, changes constantly and differs by region. We mitigated the influence of this uncertainty by developing a mean diet formulation per species based on multiple diets representative of the same period. However, still some bias for a specific region could be present and the diets used may not completely reflect global practise. Assessing diets by region and over time lay beyond the scope of this study, but doing so would provide valuable knowledge in the transition towards circular food systems. Second, the feed quantity and growth rate were not always available for each life stage of the aquaculture species selected, thus FCR could not be estimated for each life stage. Consequently, we estimated HePCR for the grow-out phase only.
As this phase has most influence on HePCR (i.e., most compound feed is consumed during it), including other phases seems unlikely to influence our conclusions. Third, we considered only protein efficiency in this study; however, because fish have a high-fat content, investigating the efficiency with which aquaculture species upgrade lipids could be a valuable next step. Overall, as an initial estimate of feed-food competition, we included multiple feeds and FCR per species, which provide a range for the HePCR. Future studies could focus on the entire life cycle of one specific system and species, as HePCR depends on the animal, its feed and efficiency and, to the definition of human edible products. 17,27,42 These types of studies could provide more detailed estimates of feed-food competition in aquaculture; nevertheless, more data should be available for these studies.
Another limitation of this study was related to its scope, as we focused only on intensive systems to the enable comparison of HePCRs between species at low versus high TLs and between aquaculture and livestock. Most aquaculture species, however, are produced in extensive or semi-intensive systems, especially finfish species at a low TL, in which they can obtain (some of) their nutritional requirements from the natural environment. As the efficiency of converting by-products is affected by species as well as production systems, 47 we need to analyse comparative case studies of specific species and production systems. In other words, we need to compare HePCRs of the same species produced in different systems and at different intensities.

| Implications of this study for future-food systems
Animals can play an important role in circular food systems by upgrading by-products; therefore, the transition towards circular systems should emphasize minimizing feed-food competition, 48,49 for both livestock and aquaculture. To ensure a net contribution of aquaculture to food security, focus should be placed not only on feed efficiency metrics, such as FCR, but also HePCR. For example, the present study, showed that Atlantic salmon, despite having a relatively low FCR, consume more protein than they produce (HePCR > 1), indicating the importance of HePCR.
In recent years, an increasing number of animal and plant-based by-products have been used as aquaculture feeds. 50,51 For example, according to recent estimates livestock by-products and fishmeal from fish by-products could replace 99% of the fishmeal made from whole fish. 9 Besides using by-products, the literature also suggests using novel protein sources to replace fishmeal, such as insects, algae and yeasts; however, the global potential to replace fishmeal with these sources, remains unclear. 52 Moreover, novel protein sources that cause feed-food competition should not be incorporated into aquaculture feeds. For example, if insects are used as fish feed they should not be fed with by-products that can be fed to the fish directly, such as slaughter waste, but rather alternative biomass streams, such as manure. A combination of novel protein sources and by-products could replace current food-competing ingredients.
To further encourage the feed industry to develop and apply innovations to increase by-product use, 11 the government could develop targets for the inclusion of by-products in aquatic feeds for feeding companies or tax the use of food-grade feed materials. 9 Alternatively, certification schemes such as the Aquaculture Stewardship A final strategy to optimize the role of aquaculture in the food system is to increase the edible yield of harvested species. If humans would consume not only fillets but also all edible parts ( The role of aquaculture in circular food systems will most likely consist of a balanced mix of species at different TLs and from different aquaculture systems, depending on the by-products available. Species at high TLs are specifically adapted to use highquality by-products from animal-based food, such as fisheries and livestock by-products, while those at lower TLs could be fed plantbased by-products or be produced in non-fed aquaculture systems.
If aquaculture species are fed mainly by-products, however, the availability and quality of these by-products would determine new boundaries for production and consumption. Hence, the consumption of aquatic species will no longer be determined by the demand, but by the amount of aquatic food that can be produced from by-products. Therefore, shifting towards circular food systems, requires changing not only production systems but also consumption patterns, especially in high-income countries. Diets should gradually contain more plant-based products and fewer animal-based products. 2 Combinations of socio-economic and institutional measures, 2 education and promotion of circular thinking, but also reduction of taxes for circular products 54 could all help achieve the change towards more circular food production.

| CONCLUSIONS
The mean natural TL in global aquaculture production has increased over time. If this trend continues, feed-food competition may also increase assuming that aquaculture feed at a higher TL contains more human edible ingredients. When considering only ingredients directly edible by the diet of the species at the highest TL, Atlantic salmon, had by far the highest human edibility. Nevertheless, in this scenario, less than 50% of each of the diets was human edible, and common carp, whiteleg shrimp and Nile tilapia were net producers of protein. When soya bean meal and fishmeal were also considered food competing ingredients, the percentage of human edibility increased substantially for all diets except that of Atlantic salmon, and all species were net consumers of protein.
Because soya bean meal is the feed ingredient most used in aquaculture, replacing it with non-food-competing protein sources is an important step towards reducing feed-food competition in aquaculture and towards circularity. The role of aquaculture in circular food systems will most likely consist of a balanced mix of species at different TLs and from different aquaculture systems, depending on the by-products available. As the natural TL is not the only factor that influences feed-food competition, future research should focus on including more species (e.g., diets, FCRs) and systems (e.g., intensities).
van RIEL ET AL.

APPENDIX
T A B L E A 1 Protein content (%), edible yield(%) and trophic level of the 50 aquaculture species most produced globally in 2019.