Aquatic Foods To Nourish Nations: Article

Article
Aquatic foods to nourish nations

https://doi.org/10.1038/s41586-021-03917-1 Christopher D. Golden1,2,3,26 ✉, J. Zachary Koehn4,26, Alon Shepon1,5,6,26, Simone Passarelli1,26,
Christopher M. Free7,8,26, Daniel F. Viana1,9,26, Holger Matthey10, Jacob G. Eurich8,11,
Received: 18 January 2021
Jessica A. Gephart12, Etienne Fluet-Chouinard13, Elizabeth A. Nyboer14 , Abigail J. Lynch15,
Accepted: 16 August 2021 Marian Kjellevold16, Sabri Bromage1, Pierre Charlebois17, Manuel Barange17,
Stefania Vannuccini17, Ling Cao18, Kristin M. Kleisner19, Eric B. Rimm1, Goodarz Danaei3,20,
Published online: 15 September 2021
Camille DeSisto21, Heather Kelahan1, Kathryn J. Fiorella22 , David C. Little23,
Check for updates Edward H. Allison24, Jessica Fanzo25 & Shakuntala H. Thilsted24
Despite contributing to healthy diets for billions of people, aquatic foods are often
undervalued as a nutritional solution because their diversity is often reduced to the
protein and energy value of a single food type (‘seafood’ or ‘fish’)1–4. Here we create a
cohesive model that unites terrestrial foods with nearly 3,000 taxa of aquatic foods to
understand the future impact of aquatic foods on human nutrition. We project two
plausible futures to 2030: a baseline scenario with moderate growth in aquatic
animal-source food (AASF) production, and a high-production scenario with a
15-million-tonne increased supply of AASFs over the business-as-usual scenario in
2030, driven largely by investment and innovation in aquaculture production. By
comparing changes in AASF consumption between the scenarios, we elucidate
geographic and demographic vulnerabilities and estimate health impacts from
diet-related causes. Globally, we find that a high-production scenario will decrease
AASF prices by 26% and increase their consumption, thereby reducing the consumption
of red and processed meats that can lead to diet-related non-communicable
diseases5,6 while also preventing approximately 166 million cases of inadequate
micronutrient intake. This finding provides a broad evidentiary basis for policy
makers and development stakeholders to capitalize on the potential of aquatic foods
to reduce food and nutrition insecurity and tackle malnutrition in all its forms.
Globally, multiple forms of malnutrition continue to be important and in improving human nutrition. The EAT–Lancet Commission report
universal. Among children under the age of five, 149 million (22%) are detailed a strategy to transform the global food system into one that
affected by stunting and 45 million by wasting7. Among adults globally, could nourish the world without exceeding planetary boundaries12. The
2.1 billion are overweight or obese8. Sparse data suggest that vitamin report, however, focused predominantly on terrestrial food produc-
A deficiency is prevalent among children in Africa and South Asia, and tion, even as it noted that it would be difficult for many populations to
zinc deficiency affects half of all children in regions for which informa- obtain adequate quantities of micronutrients from plant-source foods
tion exists9. Dietary inadequacies could be the leading reason that alone. Yet the treatment of AASFs as a homogenous group (‘seafood’
people experience multiple nutrient deficiencies and subsequent mor- or ‘fish’) has limited the potential for their inclusion and recognition
bidity and mortality10. Cardiovascular diseases, which are largely driven in global diets.
by diet-related factors, are the greatest contributor to global mortality,
causing 17.8 million deaths in 201711—greater than the approximately
2 million deaths that were caused by COVID-19 in 2020. Nutrition from aquatic food diversity
To address these multiple forms of malnutrition, contemporary food Here we reframe the role of aquatic foods in global food systems as
policy discourses centre on the role of sustainable and healthy diets a highly diverse food group, which can supply critical nutrients1–3,13
1
Department of Nutrition, Harvard T. H. Chan School of Public Health, Boston, MA, USA. 2Department of Environmental Health, Harvard T. H. Chan School of Public Health, Boston, MA, USA.
3
Department of Global Health and Population, Harvard T. H. Chan School of Public Health, Boston, MA, USA. 4Center for Ocean Solutions, Stanford University, Stanford, CA, USA. 5Department
of Environmental Studies, The Porter School of the Environment and Earth Sciences, Tel Aviv University, Tel Aviv, Israel. 6The Steinhardt Museum of Natural History, Tel Aviv University, Tel Aviv,
Israel. 7Bren School of Environmental Science and Management, University of California, Santa Barbara, Santa Barbara, CA, USA. 8Marine Sciences Institute, University of California, Santa
Barbara, CA, USA. 9Betty and Gordon Moore Center for Science, Conservation International, Arlington, VA, USA. 10Markets and Trade Division, Food and Agriculture Organization of the United
Nations (FAO), Rome, Italy. 11Department of Ecology, Evolution and Marine Biology, University of California, Santa Barbara, Santa Barbara, CA, USA. 12Department of Environmental Science,
American University, Washington, DC, USA. 13Department of Earth System Science, Stanford University, Stanford, CA, USA. 14Department of Biology, Carleton University, Ottawa, Ontario,
Canada. 15U.S. Geological Survey, National Climate Adaptation Science Center, Reston, VA, USA. 16Institute of Marine Research, Bergen, Norway. 17Fisheries and Aquaculture Division, Food and
Agriculture Organization of the United Nations (FAO), Rome, Italy. 18School of Oceanography, Shanghai Jiao Tong University, Shanghai, China. 19Environmental Defense Fund, New York, NY,
USA. 20Department of Epidemiology, Harvard T. H. Chan School of Public Health, Boston, MA, USA. 21Nicholas School of the Environment, Duke University, Durham, NC, USA. 22Department of
Population Medicine and Diagnostic Sciences and Master of Public Health Program, Cornell University, Ithaca, NY, USA. 23Institute of Aquaculture, University of Stirling, Stirling, UK. 24WorldFish,
Bayan Lepas, Malaysia. 25Bloomberg School of Public Health and Nitze School of Advanced International Studies, Johns Hopkins University, Washington, DC, USA. 26These authors contributed
equally: Christopher D. Golden, J. Zachary Koehn, Alon Shepon, Simone Passarelli, Christopher M. Free, Daniel F. Viana. ✉e-mail: golden@hsph.harvard.edu
Nature | Vol 598 | 14 October 2021 | 315

Article
and improve overall health14. Aquatic foods are defined as animals, Small pelagics
plants and microorganisms, as well as cell- and plant-based foods of
aquatic origin emerging from new technologies15. They include finfish, Clams,
mussels, oysters
crustaceans (such as crabs and shrimp), cephalopods (octopus and
squids), other molluscs (clams, cockles and sea snails), aquatic plants Large pelagics
(water spinach; Ipomoea aquatica), algae (seaweed) and other aquatic

Aquatic mammals
animals (mammals, insects and sea cucumbers). Aquatic foods can be
farmed or wild-caught, and are sourced from inland (for example, lakes, Salmonids
rivers and wetlands), coastal (estuaries, mangroves and near-shore)
and marine waters, producing a diversity of foods across all seasons Carps
and geographic regions. Here we focus on AASFs, which constitute the

Cephalopods
majority of aquatic foods.
Relative to the limited variation in domesticated terrestrial Beef
animal-source foods (for example beef, poultry, pork), AASFs pre-
sent myriad options for supplying nutrients (Fig. 1). Currently, wild Lamb
fisheries harvest more than 2,370 taxa and aquaculture growers farm Tilapia and
approximately 624 species or species-types16. To provide evidence of the other cichlids
variability in nutrient composition across this diverse array of aquatic Crab, lobster,
crayfish, shrimp
foods, we created the Aquatic Foods Composition Database17 (AFCD)
(Methods), a comprehensive global database that comprises hundreds Cods
of nutrients, including minerals (for example, calcium, iron and zinc),

Goat
vitamins and fatty acids from 3,753 aquatic food taxa. Our analysis indi-
cates that the top 7 categories of nutrient-rich animal-source foods are Chicken
all aquatic foods, including pelagic fish, bivalves and salmonids (Fig. 1).
Pork
Aquatic foods to benefit human health Omega-3

(DHA+EPA)
Vitamin
A
Vitamin
B12
Calcium Iodine Iron Zinc
AASFs improve human health through at least three pathways: by

reducing micronutrient (for example, vitamin A, calcium and iron) Per cent of recommended nutrient intake
0 50 100 or greater
deficiencies that can lead to subsequent disease; by providing the
Land food
dominant source of the omega-3 long-chain polyunsaturated fatty
acids docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA) Aquatic food
(hereafter referred to jointly as DHA+EPA), which may reduce the risk Fig. 1 | Nutrient diversity of aquatic animal-source foods in relation to
of heart disease and promote brain and eye health; and by displac- terrestrial animal-source foods. Aquatic (blue) and terrestrial (green) food
ing the consumption of less-healthy red and processed meats that richness assessed as a ratio of concentrations of each nutrient per 100 g to the
can cause adverse health outcomes14. Any of these three pathways daily recommended nutrient intake. Each shaded box represents the median
may overlap in an individual, or predominantly target consumers of value of each nutrient in a muscle tissue across all species within each
particular geographies or age–sex groups. The third pathway, spe- taxonomic group. Food groups were ordered vertically by their mean nutrient
cifically, is characteristic of the nutrition transition—the process by richness with higher values meeting a higher percentage of the daily
which demographic and economic shifts lead to concomitant dietary recommended intake. See Supplementary Table 4 for the recommended
and epidemiological shifts that often accompany the Westernization nutrient intake values and their citations.
of food systems18. To better understand these pathways, we examine
how aquatic food policy initiatives and investments could improve
diets and public health through increasing access to the diversity of second, a high-AASF-production scenario that assumes higher growth
aquatic foods and the nutrients that they provide. rates in production as a result of increased financial investment and
We explicitly integrated aquatic and terrestrial food-systems models innovation in aquaculture and improved and effective management in
to evaluate the potential health impacts of increasing global AASF capture fisheries21 (Methods). The projections are not forecasts about
production. This integration enables a more realistic portrayal of the the future, but rather plausible scenarios based on a set of internally
trade-offs made within our global terrestrial and aquatic food systems. consistent assumptions. Increases in aquaculture and capture fisheries
To understand the health impacts of increased AASF consumption, we in the high-production scenario led to a 26% decrease in the interna-
modelled future food systems to 2030. We used an integrated version tional reference price of AASFs, and an increase in their production
of the FISH model19 from the United Nations Food and Agriculture by 15.5 million tonnes (an approximate 8% increase in annual global
Organization (FAO), and the Aglink–Cosimo model20, which is jointly production) in 2030 as compared to the baseline scenario. In each
maintained by the Organization for Economic Cooperation and Devel- scenario, we calculated the nutrients supplied to 191 countries from the
opment (OECD) and the FAO. The embedded budgeting framework projected composition of the food-system models by assigning nutri-
and price elasticities across foods enabled the addition of AASFs and ent composition values to the suite of foods being consumed within
the substitution of aquatic for terrestrial foods within national diets. 22 food commodity categories, using the Global Nutrient Database
This affects the supply and demand of a broad range of related food (GND)22 and the AFCD. For 21 of the 22 food commodity categories
items, particularly terrestrial animal-source foods (such as poultry, (all terrestrially produced foods), the GND was used as the source of
pork, beef, eggs and dairy). nutrient composition data. For the one commodity category containing
We used the integrated model to produce two scenarios: first, a base- aquatic foods, the AFCD nutrient composition values were used. A set of
line scenario with projections of moderate growth trends in AASF pro- refuse factors is applied to all foods, highly specific to individual foods
duction and expert consensus regarding macroeconomic conditions, and their respective forms of preparation. Within the food group of fish
agriculture and trade policy settings, long-term productivity, inter- and seafood, these refuse factors vary from 55% for fresh crustaceans
national market developments and average weather conditions; and to 10% for fresh cephalopods.
316 | Nature | Vol 598 | 14 October 2021

a Aquatic animal-source food consumption b Red meat consumption
Difference in Difference in
2030 intake 2030 intake
(high vs base) (%) (high vs base) (%)
20
0
15
–1
10
5 –2
–3
c Poultry consumption d Egg consumption
0
0
–1
–1
–2
–2
–3
–3
e Dairy consumption f Non-aquatic animal-source food consumption
0.1
0
0
–0.1 –1
–0.2
–2
–0.3
Fig. 2 | Shifts in fish and red meat consumption resulting from an increase high-production and baseline-production scenarios. Values greater than zero
in aquatic animal-source foods. a–f, The percentage difference in indicate greater consumption under the high-production scenario. Countries
consumption of mean aquatic animal-source food (a), red meat (bovine, ovine smaller than 25,000 km2 are illustrated as points (small European countries
and pork) (b), poultry (c), egg (d), dairy (milk, butter and other dairy products) excluded). All European Union member countries have the same value because
(e) and all non-aquatic animal-source food (f) in 2030 under the they are modelled as a single unit in the Aglink–Cosimo model.
To assess the role of diversity in the aquatic food system, we com- disease and certain types of cancer5,6. Second, AASFs displace the
pared estimated nutrient outputs with and without species diversity consumption of more harmful animal-source foods—such as red and
fully disaggregated at national levels. The GND uses relatively similar processed meats (Fig. 2, Extended Data Figs. 2–4, Supplementary
nutrient composition values across all aquatic foods, varying only for Data 1)—particularly in the global north, or can attenuate their increased
the 12 categories explicitly modelled in the GND (for example, demersal consumption in the global south23,24, in both cases reducing the risk of
fish, pelagic fish and so on). We disaggregated national consumption diet-related non-communicable disease25.
to the species level in proportion to species-specific aquaculture and In much of the global north, an increase in AASF consumption was
capture-fisheries production reported by the FAO, and linked these associated either with reductions in the consumption of red meat,
disaggregated species to the AFCD (Methods). Instead of 12 GND poultry, eggs and dairy, or with no notable impact (that is, no discern-
categories for aquatic foods, we used supply and nutrient composi- ible increases; Fig. 2). In the global south, an increase in AASF consump-
tion values for 2,143 taxa. This comparison enabled us to determine tion was not associated with declines in the consumption of red meat,
whether incorporating species diversity, as opposed to relying on poultry, eggs and dairy. The combined dietary effect of increasing
common commercial species, shifted the levels of nutrients supplied AASFs and reducing red and processed meats may lead to a reduced
by aquatic foods. The disaggregated model outputs in the baseline risk of hypertension, stroke, heart disease, diabetes, colorectal cancer
scenario resulted in a higher supply of calcium (8% higher; median and breast cancer. Countries that are rapidly undergoing the nutrition
across countries), iron (4%), DHA+EPA (186%), zinc (4%) and vitamin B12 transition (such as China, India, Philippines, Malaysia, Indonesia, Viet-
(13%), with a 1% decline in vitamin A (Extended Data Fig. 1). This result nam, South Korea, Mexico, Brazil, Peru, Chile, Nigeria, Russia, USA and
provides evidence that narrowly focusing on the nutrient contributions Canada) are most likely to benefit from increases in AASF production,
of commercially important species underestimates the nutritional which could avert the trajectory of their populations towards harmful
benefits of aquatic foods, especially from diverse small-scale fisheries. levels of meat consumption (Fig. 2).
Aquatic foods can reduce meat intake Aquatic foods can fill the nutrient gap
In addition to the key role of AASFs in providing essential micronu- Deficiencies in key micronutrients—such as iron, zinc, calcium,
trients, DHA+EPA and protein, AASFs can also prevent diet-related iodine, folate, vitamins A, B12 and D—have led to 1 million premature
non-communicable diseases. These health benefits are delivered deaths annually8. Further, an estimated 30% of the global population
through two mechanisms. First, AASFs directly provide DHA+EPA, (around 2.3 billion people) have diets that are deficient in at least one
which may improve brain function and reduce the incidence of heart micronutrient8. Inadequate nutrient intakes can arise from various

Article
DHA+EPA Vitamin B12
ΔSEVs (%) ΔSEVs (%)

(high – base) (high – base)
0
0
–1
–0.25
–2
–3 –0.50
–4 –0.75
≤–5 ≤–1.0
Iron Zinc

0 0.2
–0.25 0
–0.2
–0.50
–0.4
≤–0.75 ≤–0.6
Calcium Vitamin A

0.5 1.0
0 0.5
–0.5 0
≤–1.0
Women and girls Men and boys ΔMillions of people

with inadequate intakes
DHA+EPA (high – base)
Vitamin B12 0
Iron –1
Zinc –2
Calcium –3
Vitamin A –4
–5
0–4
5–9
10–14
15–19
20–24
25–29
30–34
35–39
40–44
45–49
50–54
55–59
60–64
65–69
70–74
75–79
80–84
85–89
90–94
95–99
0–4
5–9
10–14
15–19
20–24
25–29
30–34
35–39
40–44
45–49
50–54
55–59
60–64
65–69
70–74
75–79
80–84
85–89
90–94
95–99
Age group
Fig. 3 | Shifts in micronutrient intake resulting from an increase of aquatic number of people with inadequate micronutrient intakes, by age–sex group.
animal-source foods. The maps show the difference in SEVs in 2030 under the Values less than zero indicate fewer inadequate intakes under the
high-production and baseline-production scenarios by country. Values less high-production scenario. Countries smaller than 25,000 km2 are illustrated as
than zero indicate reduced risk (lower SEVs) of inadequate intake under the points (small European countries excluded).
high-production scenarios. The bottom panels show the difference in the
factors: the formulation of food systems, including the availability taxa included in the analysis, we used standardized nutrient com-
and accessibility of foods; ecological or environmental conditions— position values for muscle tissue because the species coverage was
such as soil nutrient loss, drought or fishery declines—that decrease higher than for other parts (such as liver, bones and eyes). Because
availability; reduced access to markets and natural resources through these other parts are often more nutrient-rich than muscle tissue,
tariffs, fisheries governance, or other economic incentives; and/or our estimates are likely to be conservative, underestimating the true
taste preferences, consumer behaviour or other individualized fac- value of AASFs in human diets.
tors8,26,27. AASFs have the capability to reduce or fill this nutrient gap We calculated summary exposure values (SEVs) to assess the excess
with bioavailable forms of micronutrients, particularly in geographies risk that each country experiences because of inadequate nutrient sup-
where AASF reliance and nutritional deficiencies are high, such as ply in their overall food systems, comparing the total amount of nutri-
equatorial regions1. tion derived from apparent consumption against age- and sex-specific
Here we focus on nutrient supply to estimate the contribution of nutrient demands (Methods). SEVs range from 0% to 100% and should
AASFs to overall nutrient intake. In the high-production scenario be viewed as a risk-weighted prevalence, with higher SEVs representing
by 2030, AASFs may contribute a global average of 2.2% of energy, higher risk of inadequate micronutrient intake28. The difference in SEVs
13.7% of protein, 8.6% of iron, 8.2% of zinc, 16.8% of calcium, 1.1% of represents the change in potential risk of inadequate nutritional intake
vitamin A, 27.8% of vitamin B12 and 98–100% of DHA+EPA, an approxi- between the two AASF production scenarios in 2030 (Fig. 3, Supplemen-
mate 0–10% increase for each nutrient above 2020 reference values tary Data 3). With overall trends in increasing AASF consumption and
(Extended Data Fig. 5, Supplementary Data 2). For each of the AASF concomitant reductions in poultry, eggs, dairy, and red and processed

meats (Fig. 2), there are large gains in micronutrient and DHA+EPA intakes of women and girls (average of 51.4% of countries) over men
consumption (Fig. 3). Globally, the high-production scenario will lead and boys (average of 18.2% of countries; Supplementary Data 4). Thus,
to reductions in inadequate intake across most assessed nutrients there is an almost three times greater likelihood of increased AASF con-
(reduction of 8.1 million iron, 5.5 million zinc, 49.3 million calcium, sumption benefitting female nutrition, providing a potential pathway
36.0 million vitamin B12, and 76.8 million DHA+EPA inadequate intakes), for nutritional equity (Supplementary Methods).
while potentially increasing 10.1 million vitamin A inadequate intakes
(Extended Data Fig. 6). Particular geographies will also experience
small declines in calcium, iron, vitamin A and zinc supply. This phe- Discussion
nomenon probably arises from modest reductions in the consump- We illustrate the role of AASFs in improving the future of human health,
tion of iron- and zinc-rich red meat (as shown in historical trends), and focusing on supplying critical micronutrients and attenuating mor-
large reductions in the consumption of calcium- and vitamin-A-rich bidity and mortality from chronic disease that is characteristic of the
dairy, egg and poultry. Notably, certain regions that are character- nutrition transition. Our analyses demonstrate that an increase in pro-
ized by food and nutrition insecurity (for example, sub-Saharan Africa duction of the rich diversity of AASFs (and the nutrients contained
and Southeast Asia) experience increases in intake for all measured therein) can improve the diets of many nations. Notably, our analysis
nutrients. However, some populations will face increasing risk of inad- focuses on the consumption of muscle tissue from AASFs and therefore
equate micronutrient intake if consumption of AASFs displaces other must be viewed as a probable underestimate of the potential contribu-
foods, as evidenced by reduced calcium intake in Turkey, zinc intake tion to micronutrient supplies. Our projection of declines in global
in Azerbaijan, and vitamin A intake in Indonesia and Mexico, among vitamin A supply may be incorrect, given the high levels of this nutrient
others (Fig. 3). Yet, globally, there is a pattern in which increasing the in certain fish parts (such as liver) that are not included because of our
diversity of aquatic animal-source food consumption leads to reduced focus on muscle tissue.
micronutrient-inadequate intakes (Extended Data Fig. 7). The diversity of aquatic foods highlighted here evidences the limita-
Recognition of the diversity of AASFs and their nutrient composition tions of treating them as a homogenous group. The EAT–Lancet Com-
could be harnessed to direct their production and consumption across mission Report12 undervalues the importance of aquatic foods; key
a range of deficient minerals, fatty acids and vitamins. For instance, if food policy dialogues (such as the UN Sustainable Development Goal
calcium deficiency is an issue in Turkey, one prudent option might be 2: Zero Hunger) ignore aquatic foods completely; and funding for the
to increase the consumption of pelagic small fish (such as herrings and aquatic foods sector from the World Bank and Regional Development
sardines)29. Similarly, if vitamin A deficiency is an issue in Brazil, then Banks lack targeted support31. Two main issues seem to be pervasive
efforts to promote the production of oysters or the consumption of in misunderstanding the importance of aquatic foods. First, a very
sardines might be appropriate30. These types of food-system solutions narrow view of the diversity of fish and seafood is often taken, with
will require sub-national targeting of vulnerable populations and will a focus on a set of commercially grown or wild-harvested finfish and
rely on efforts to increase both production and consumption. bivalves. This classification ignores the vast diversity of other spe-
cies, forms of culture production, and wild harvest by small-scale
fisheries32. Second, the nutritional contribution of aquatic foods has
Aquatic foods support the vulnerable traditionally focused on its low contribution to global energy (that
Diets are shaped by the structure of food systems. Access to the foods is, calories) and protein intake, failing to consider the contribution
produced by these systems can vary by age, sex, culture, socio-economic of aquatic foods to nutrition via highly bioavailable essential micro-
status and geography, as does a given population’s reliance on AASFs. nutrients and fatty acids. The AFCD presented here enables future
AASFs are important for both sexes and all ages, but particularly so for studies to move beyond this limited view of nutrition from aquatic
young children, pregnant women and women of childbearing age, due foods. However, there are still limitations in our current presenta-
to the critical role of micronutrients and DHA+EPA in fetal and child tion (for example, a lack of focus on vitamin D due to variable intake
growth and development30. requirements and a lack of recognition of the nutritional value of
Because different age–sex groups have different vulnerabilities small fish and non-muscle fish parts in human nutrition). Vitamin D
to certain health outcomes, a disproportionate benefit is associated deficiency is a major health issue in some countries, and an increase
with consuming AASFs for particular groups. The function of reduc- in fatty fish intake could reduce this.
ing micronutrient deficiencies is more important for children and It is critical to consider where and how aquatic foods are produced,
women of reproductive age, and the function of attenuating morbid- because environmental, social and economic impacts can vary widely
ity and mortality as a result of chronic disease is more important for across both the wild-capture and aquaculture sectors (Supplementary
adults. For example, older people in Tunisia, Algeria, St Lucia, Iran Methods). Despite the variability in environmental impacts across
and Moldova would experience large benefits in reduced inadequate animal-source food-production sectors, aquaculture and wild-capture
intake of DHA+EPA (ΔSEV of at least −10.0 percentage points) and fisheries nearly always produce fewer greenhouse gas emissions and
reduced inadequate intake in iron in Kiribati and the Republic of the use less land than the farming of red meats, and many AASFs outper-
Congo (ΔSEV = −3.6 percentage points). In several countries, children form poultry33. Sustainably and equitably achieving the human health
would experience large benefits in reduced inadequate calcium intake benefits of expanded aquatic food production will require policies and
due to increased AASF consumption (ΔSEV percentage points for technologies that mitigate impacts on adjacent ecosystems, industries
5–9-year-olds = −6.0 for girls and −5.5 for boys in Myanmar; −5.9 for girls and communities21.
in Vietnam and Cambodia; −5.1 for girls in Morocco; and −4.5 for boys
and girls in Gabon; and ΔSEV percentage points for 0–4 year-olds = −4.9
for girls and −4.4 for boys in Maldives and −4.7 for boys and −4.3 for girls Policy translation
in Kiribati). In Panama, Iran, Moldova, Dominica and Egypt, a segment Our findings suggest the following strategic research and policy oppor-
of reproductive-aged women (25–49 years) would receive a large health tunities:
benefit from increased DHA+EPA consumption (ΔSEV= −6.7 to −8.6 per- First, in countries in which there are high burdens of micronutrient
centage points). Across all measured nutrients, there were significant deficiencies, the supply chains and availability of aquatic foods may be
sex differences in benefits between the base and the high-production strengthened by improving fisheries management; enhancing sustain-
scenario (n = 73 of a total 115 age–nutrient groups), in which increased able aquaculture; and building more equitable national and regional
AASF production and consumption disproportionately improved the trade networks.

Article
Second, the promotion of a diversity of nutrient-rich aquatic foods 9. Victora, C. G. et al. Revisiting maternal and child undernutrition in low-income and
middle-income countries: variable progress towards an unfinished agenda. Lancet 397,
in sustainable aquaculture systems, in designing national food-based 1388–1399 (2021).
dietary guidelines, and for public-health interventions targeting par- 10. Afshin, A. et al. Health effects of dietary risks in 195 countries, 1990–2017: a systematic
ticular nutritional deficiencies among vulnerable populations living analysis for the Global Burden of Disease Study 2017. Lancet 393, 1958–1972 (2019).
11. Mensah, G. A., Roth, G. A. & Fuster, V. The global burden of cardiovascular diseases and
in particular geographies. risk factors: 2020 and beyond. J. Am. Coll. Cardiol. 74, 2529–2532 (2019).
Third, incentivizing access and affordability of aquatic foods in 12. Willett, W. et al. Food in the Anthropocene: the EAT–Lancet Commission on healthy diets
countries experiencing a rapid nutrition transition. from sustainable food systems. Lancet 393, 447–492 (2019).
13. Youn, S. J. et al. Inland capture fishery contributions to global food security and threats to
Fourth, prioritizing aquatic foods in social protection programs, their future. Glob. Food Sec. 3, 142–148 (2014).
including food assistance, school meal programmes, and safety nets 14. Rimm, E. B. et al. Seafood long-chain n−3 polyunsaturated fatty acids and cardiovascular
for the most nutritionally vulnerable, including pregnant and lactating disease: a science advisory from the American Heart Association. Circulation 138, e35–
e47 (2018).
women, young children in the first 1,000 days of life, and older people. 15. WorldFish. 2030 Research and Innovation Strategy: Aquatic Foods for Healthy People and
In line with the Voluntary Guidelines on Food Systems and Nutrition34 Planet (eds Lala-Pritchard, T. & Johnstone, G) (WorldFish, 2020).
of the Committee on World Food Security, national food and nutrition 16. Global Production by Production Source 1950–2019. FishStatJ (FAO, 2021).
17. Golden, C. D. et al. Aquatic Food Composition Database https://doi.org/10.7910/DVN/
policy is needed to transform food systems by prioritizing aquatic foods KI0NYM (Harvard Dataverse, 2021).
where culturally and socially appropriate. Also, policy may ensure that 18. Popkin, B. M. & Gordon-Larsen, P. The nutrition transition: worldwide obesity dynamics
the governance of and investment in aquatic food systems aims to pre- and their determinants. Int. J. Obes. 28, S2–S9 (2004).
19. The State of World Fisheries and Aquaculture 2020. 186–193 (FAO, 2020).
serve, support and improve aquatic species diversity; production and 20. OECD/FAO. OECD-FAO Agricultural Outlook 2021–2030. https://doi.org/10.1787/
harvest methods and practices; and efficient and safe distribution chan- 19428846-en (OECD Publishing, 2020).
nels. With more than 1.5 billion people unable to afford a healthy and 21. Costello, C. et al. The future of food from the sea. Nature 588, 95–100 (2020).
22. Schmidhuber, J. et al. The Global Nutrient Database: availability of macronutrients and
sustainable diet35, our model results showcase the importance of price micronutrients in 195 countries from 1980 to 2013. Lancet Planet. Health 2, e353–e368
and economic policies in creating nutritious diets that are affordable (2018).
for consumers. These measures should enable aquatic foods to have 23. Dey, M. M. et al. Demand for fish in Asia: a cross-country analysis. Aust. J. Agric. Resour.
Econ. 52, 321–338 (2008).
an important role in nourishing the global population and improving 24. Gallet, C. A. The demand for fish: a meta-analysis of the own-price elasticity. Aquac. Econ.
global nutrition and health. Manag. 13, 235–245 (2009).
25. Zhao, L. G. et al. Fish consumption and all-cause mortality: a meta-analysis of cohort
studies. Eur. J. Clin. Nutr. 70, 155–161 (2015).
26. Vermeulen, S. J., Park, T., Khoury, C. K. & Béné, C. Changing diets and the transformation
Online content of the global food system. Ann. NY Acad. Sci. 1478, 3–17 (2020).
27. Naylor, R. L., et al. The demand for blue foods across geographic and temporal scales.
Any methods, additional references, Nature Research reporting sum-
Nat. Commun. (in the press).
maries, source data, extended data, supplementary information, 28. Murray, C. J. L. et al. Global burden of 87 risk factors in 204 countries and territories,
acknowledgements, peer review information; details of author con- 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet
396, 1223–1249 (2020).
tributions and competing interests; and statements of data and code
29. Isaacs, M. The humble sardine (small pelagics): fish as food or fodder. Agric. Food Secur.
availability are available at https://doi.org/10.1038/s41586-021-03917-1. 5, 27 (2016).
30. Bernstein, A. S., Oken, E. & de Ferranti, S. Fish, shellfish, and children’s health: an
1. Golden, C. D. et al. Nutrition: fall in fish catch threatens human health. Nature 534, assessment of benefits, risks, and sustainability. Pediatrics 143, e20190999 (2019).
317–320 (2016). 31. Bennett, A. et al. Recognize fish as food in policy discourse and development funding.
2. Byrd, K. A., Thilsted, S. H. & Fiorella, K. J. Fish nutrient composition: a review of global Ambio 50, 981–989 (2021).
data from poorly assessed inland and marine species. Public Health Nutr. 24, 476–486 32. Gelcich, S. et al. Challenges and opportunities for small-scale actors in aquatic food
(2021). systems. Nature Food (in the press).
3. Hicks, C. C. et al. Harnessing global fisheries to tackle micronutrient deficiencies. Nature 33. Gephart, J. A. et al. Environmental performance of blue foods. Nature https://doi.org/
574, 95–98 (2019). 10.1038/s41586-021-03889-2 (2021).
4. Bernhardt, J. R., & O’Connor, M. I. Aquatic biodiversity enhances multiple nutritional 34. Hirvonen, K., Bai, Y., Headey, D. & Masters, W. A. Affordability of the EAT–Lancet reference
benefits to humans. Proc. Natl Acad. Sci. 118, e1917487118 (2021). diet: a global analysis. Lancet Glob. Health 8, e59–e66 (2020).
5. Manson, J. E. et al. Marine n−3 fatty acids and prevention of cardiovascular disease and 35. Committee on World Food Security. The CFS Voluntary Guidelines on Food Systems and
cancer. N. Engl. J. Med. 380, 23–32 (2019). Nutrition (VGFSyN) CFS 2021/47/7/Rev.1 (FAO, 2021).
6. Meat, Fish and Dairy Products and the Risk of Cancer. Continuous Update Project Expert
Report 2018 (World Cancer Research Fund/American Institute for Cancer Research, 2018). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
7. UNICEF, WHO & The World Bank Group. Levels and Trends in Child Malnutrition. (WHO, 2021). published maps and institutional affiliations.
8. Global Nutrition Report: Action on Equity to End Malnutrition (Development Initiatives
Poverty Research, 2020). © The Author(s), under exclusive licence to Springer Nature Limited 2021

Methods a wholesale price) and the GDP deflator to capture movement in the
other costs along the supply chain. The higher the GDP of a country,
Food system modelling approach the smaller the influence of the wholesale price in the calculation of
The FAO FISH19 and Aglink–Cosimo20 models are recursive-dynamic, the retail price. Imports and exports are a function of the ratio between
partial equilibrium models used to simulate developments of annual the domestic (adjusted by tariff and exchange rate) and world price of
market balances and prices for the main agricultural commodities aquatic products with different levels of responsiveness depending
produced, consumed and traded worldwide. Aglink–Cosimo and FAO on the openness of the different countries’ aquatic product markets.
FISH are managed by the Secretariats of the OECD and FAO, and used Finally, the price of traded aquatic products is the market clearing
to generate the annual OECD–FAO Agricultural Outlook20 and other variable of each country component.
peer-reviewed scenario analyses36. The references cited provide full
model descriptions. Scenario development
The FAO FISH model contains 2,019 equations and covers 47 country Two alternative outlook projections, a baseline and high-production
and/or region endogenous modules. Three products are covered with scenario (Supplementary Table 1; Supplementary Fig. 1), were used to
complete supply-disposition variables and prices: an aggregate of all represent food production, consumption, and trade to 2030 for 22 food
aquatic animals except mammals; fishmeal; and fish oil. For the aggre- groups. The baseline scenario is driven by the results of the FAO FISH
gate aquatic animals, the model supplies functions for both capture model included in the OECD–FAO Agricultural Outlook 2020–2029,
and aquaculture depending on the country or regional aggregate. On with 2030 data reflecting the UN FAO’s best understanding of likely
the demand side, the model produces one aggregate aquatic animal fisheries and aquaculture growth (Supplementary Fig. 2) based on
demand function, but includes 3 different types of use: food; processed anticipated macroeconomic conditions, agriculture and trade policy
into fishmeal and oil; and other uses. settings, fisheries management outcomes, long-term productivity,
To reflect the fact that fisheries are a renewable natural resource that international market developments and average weather conditions37.
are fully exploited and regulated or over-exploited, capture fisheries Aquaculture will be the main driver of the growth up to 2030, while fish-
are kept exogenous in most modules of the model as they are controlled eries production is expected to slightly decline. The high-production
under strict fishing quotas and subject to regulations preventing eco- scenario is not a prediction but represents the UN FAO’s specific estima-
nomically driven supply. Therefore, the supply of only 11% of world tion of the upper limits of aquatic foods growth potential37, reflecting
capture fisheries respond to price for those countries and regions an imposed change to AASF production. This could be obtained by
with insufficiently strict regulations. However, it is assumed that their applying innovative technologies, capacity building, increased and
capture production will always stay below the maximum sustainable cost-effective financial investment in aquaculture and improved and
yield. Conversely, in the model, 99% of world aquaculture produc- effective management in fisheries production constrained by estimates
tion is endogenous and responsive to the price of the output, and 75% of global maximum sustainable yield. Also in the high-production sce-
of aquaculture is additionally responsive to feed prices. In terms of nario, major growth in production is expected to originate mainly from
aquaculture supply, the model contains 115 functions that cover the aquaculture, but fisheries production will slightly grow. The improved
combination of countries and species. Each species has its specific and effective management will support the sustainable growth in fish-
feed rations (different mix of feed ingredients), production lags driven eries production through increased catches in areas recovering from
by the species biology, and elasticities (the level of responsiveness previous overexploitation patterns, as well as underfished resources,
of production to price changes). Ninety seven per cent of the global and improved utilization of the harvest, including reduced onboard
reduction of fish into fishmeal and oil is endogenous in the model. In discards, waste and losses.
63% of the modules, fishmeal and oil is controlled by a simple technical Although the high-production scenario is optimistic, it is within the
parameter, whereas in the remaining modules it is price-responsive. realm of possible futures, and is used to explicitly highlight the potential
The Aglink–Cosimo model, described as a structural sector model, nutritional and health gains that could arise from targeted interven-
provides a mathematical representation of the decision processes of tions. Species composition of broad commodity categories and feed
producers and consumers of agricultural commodities. The equations composition (which could affect nutrient composition of products)
relate exogenously provided projections of the macroeconomic envi- were left unchanged between the present and 2030. We estimated
ronment, such as population growth and gross domestic product (GDP) country-level AASF consumption corresponding to marine and fresh-
developments, through commodity- and country-specific parameters water capture and aquaculture production projections in 2030 based
to agricultural supply and demand variables. These variables are pro- on the joint Aglink–Cosimo FISH baseline and high-production outputs.
jected forward in a dynamic-recursive way using prices at domestic and As the supply of fish is increased relative to the baseline, under
global levels to clear markets at all stages. The demand for food is a func- the assumption that demand does not shift, a new equilibrium price
tion of income, own and cross prices, in which the respective elasticities is found along the demand curve. This new price of fish influences
control the relative strength of each variable. Because Aglink–Cosimo the consumption and production of other agricultural commodities
and the FAO FISH model are ‘partial-equilibrium’ sector models, income through links on the production and consumption side. The shift in the
does not change in the scenario. The substitution between the various international reference price of fish, which represents the aggregate
food items is caused by shifts in relative prices. behavior of all consumers, leads to changes in individual decisions that
The FAO FISH model was integrated into Aglink–Cosimo to represent are determined by the relative changes in their domestic prices. They,
the aquatic foods component of the overall global food and agricul- in turn, are determined by the integration of each commodity market
ture system. Once integrated, the fish, fishmeal and fish oil of the FISH into the global trade system and the respective shift of the fish supply
model become fully integrated into the merged model and the full set in the scenario. Consumers in a fish-producing or importing country
of commodities is simulated simultaneously. Per capita food demand will take advantage of the lower fish price and consume more fish and
of aquatic products is determined by their retail price, retail price of less terrestrial meats, depressing terrestrial meat prices. These prices
substitutes (mostly beef, pork and poultry), and by real per capita GDP. are also transmitted through trade to countries that do not produce or
Typically, consumers from wealthier countries respond less to a change import a substantial volume of fish. Thus, consumers take advantage
in the retail price of fish expressed in real terms (that is, deflated by the of the lower meat prices and increase their meat consumption.
overall consumer price index) than consumers who spend a higher On the production side, similar effects are simulated. As demand
share of their income on food. The retail price of aquatic products is for meat declines globally owing to its substitution with cheaper fish,
determined by the price of traded products (which can be considered demand for feed also declines, lowering its price. Depending on the
Article
production technology, certain producers take advantage of the were sourced from international food composition databases from the
cheaper feed and increase production of livestock products. As cere- United States Department of Agriculture (USDA), FAO INFOODS and
als are used as feed and food, the consumption of staples also increases. the EU SMILING project in Southeast Asia, as well as individual food
The relative size of all of these responses culminates in the trade flows. composition tables from Australia, New Zealand, Pacific Islands, South
They shift relative to the baseline and a new global market equilibrium Korea, India, Bangladesh, West Africa, Canada, Norway and Hawaii, and
is found. A full description of the high-production scenario parameters previous reviews of peer-reviewed literature2.
and assumptions can be found in the Supplementary Methods. The search strategy focused on studies between 1990 and 2020, and
prioritized specific journals known to include food composition data
Global Nutrient Database (for example, Food Chemistry, Journal of Food Composition and Analysis).
The GND matched over 400 food and agricultural commodities from A broader search was also conducted using Web of Science including
the FAO’s Supply and Utilization Accounts to food items in the United 20 aquatic and 15 nutritional search terms, with elimination hedges to
States Department of Agriculture Food Composition Database and avoid irrelevant studies (see Supplementary Methods for full terms).
obtained data on nutrient composition of the Supply and Utilization Peer-reviewed data were collected from 1,063 individual studies. In total,
Accounts food items22. After adjusting for the inedible portion of each the AFCD contains 29,912 lines of data representing 3,753 unique taxa.
food item, the GND can estimate the national availability of macro- We estimated the likely mix of species consumed as described above
nutrients and micronutrients in a given year. On the basis of this, the and then matched these individual species identities with the AFCD. To
22 food group model outputs from the Aglink–Cosimo model were link disaggregated species to the AFCD, we used a hierarchical approach
cross-walked to the GND, and nutrient supply was estimated for each to assign the nutritional value for all 7 nutrients to all species consumed
scenario (Supplementary Table 1). globally (Supplementary Fig. 5). When multiple entries were present
for a single species, we took the mean of all entries. We built this hier-
Species disaggregation archy according to the following order: scientific name, average of
Aquatic foods in the GND are based on FAO FishStat production data species genus, average of species family, common name, average of
and currently include the following categories: demersal fish; pelagic species order, and average of GND category. In the disaggregation
fish; fish oils; crustaceans; cephalopods; other marine fish; freshwater effort, we found 2,143 different aquatic species being consumed glob-
fish; other molluscs; aquatic mammals; other aquatic animals; and ally. We matched nutrient composition values from muscle tissue for
aquatic plants. To derive more resolved consumption estimates, we protein, iron, zinc, calcium, vitamin A, vitamin B12 and DHA+EPA. After
first assigned fish consumption estimates to freshwater and marine this matching process, we updated the estimates of nutrient intake at
species on the basis of historical shares. Within these broad categories, national levels.
consumption was then assigned to capture and aquaculture sources
to allow for future projections to reflect increased share (for some key Sub-national intake distributions
species) in aquaculture production. Next, we used FAO FishStat produc- To evaluate the health impacts of AASF consumption, we first mod-
tion data to predict which species are actually being consumed in each elled the distribution of habitual dietary intake across age–sex groups
country, adjusting for trade flows. We assumed that future diets pre- and geographies. Using SPADE (Statistical Program to Assess Habitual
served the current taxonomic make-up within each of these categories. Dietary Exposure), an R-base package that uses 24-hour recall data to
For marine species disaggregation, we used country-specific FAO remove within-person variability and estimate habitual intake distri-
FishStat historical catch and production data from 2014 to proportion- butions39, we estimated usual intakes of iron, zinc, calcium, vitamin
ally assign consumption projections to the Aglink–Cosimo outputs. A, vitamin B12, DHA+EPA and red meat. These distributions relied on
Freshwater species, with the exception of salmon (calculated separately the availability of individual dietary intake data with variable days of
using FAO trade data), and any fish destined to fishmeal, fish oil or dis- 24-hour recalls, which were available in 13 datasets to which we had
cards were removed. National apparent consumption of marine seafood access, including: United States, Zambia, Mexico, China, Lao PDR, Phil-
by species from all producing sectors and sources (aquaculture, capture ippines, Uganda, Burkina Faso, Bulgaria, Romania, Italy, Bangladesh and
and import) was calculated by subtracting exports from production, Bolivia. A summary of the datasets used to estimate the sub-national
using FAO food balance sheets (according to the proportion of species intake distributions is available in Supplementary Table 7.
within each seafood commodity category), and adding imports (assum- We fit gamma and log-normal distributions to the habitual intake
ing a species mix within trade codes proportional to trade partner distributions for all available age–sex groups using the fitdistrplus pack-
production). Negative apparent consumption was assumed to be zero. age40. We selected the distribution with the best Kolmogorov–Smirnov
Finally, we scaled total harvest by the edible portion of each species. (KS) goodness-of-fit statistic (0.002–0.373) as the final distribution for
Consumption of freshwater taxa was generated by matching FAO each group. The parameters of this best fitting distribution describe
FishStat production and trade labels nested in the same commodity the shape of habitual intake distribution for each age–sex group and
group (Supplementary Methods; Supplementary Figs. 3, 4). All com- can be shifted along the x axis in response to changing diets.
modities were converted to live weights using freshwater conversion
factors38. The proportion of freshwater species consumed was further Assigning national intake distributions
disaggregated with household survey data38, and recreational fishery We disaggregated country-level intakes into sub-national distributions
consumption (Supplementary Methods). Household surveys were of intake in three steps. First, we disaggregated the European Union,
used to adjust the volume of capture fishery relative to aquaculture which is modelled as a single entity in the integrated model, into its
in 31 countries and disaggregated unidentified commodity groups for 27 constituent countries (Supplementary Table 5). Second, we disag-
five countries38. Recreational fisheries data from ancillary sources were gregated country-level mean intakes into age–sex-level mean intakes
included for 11 countries that have high but potentially under-reported using the Global Expanded Nutrient Supply (GENuS) database41 for
recreational participation. Finally, we estimated consumable harvest by all nutrients except DHA+EPA and vitamin B12, which are not included
scaling total harvest by edible proportion (Supplementary Methods). in the GENuS database. We used the SPADE habitual intake output to
derive age–sex-level mean intakes for these two nutrients. Finally, we
Aquatic Foods Composition Database used the SPADE habitual intake output to describe the shape of intake
The AFCD synthesizes information from international and national distribution for each age–sex group.
food composition tables and peer-reviewed literature. Food composi- The GENuS database uses historical national dietary trend data to
tion tables were assumed to be correct and directly integrated. Data estimate the availability of 23 individual nutrients across 225 food
categories for 34 age–sex groups in nearly all countries in 201141. We proxy zinc and iron values, based on its social development index50.
used these estimates to calculate scalars for relating country-level For vitamin B12, we used the values used by the Institute of Medicine51
availability to age-group-level availability as: but acknowledge that uncertainties regarding recommended intakes
exist, and used a coefficient of variation of 25% instead of the default
Scalarc , n, a, s = availabilityc , n, a, s/mean(availabilityc , n) 10% in constructing our risk curves52.
where the scalar for country c, nutrient n, age group a and sex s is cal- Reporting summary
culated by dividing the nutrient availability for each age–sex group by Further information on research design is available in the Nature
the mean nutrient availability for all age–sex groups. We assume these Research Reporting Summary linked to this paper.
ratios of nutrient availability are proportional to ratios of nutrient
intake and scale the country-level mean nutrient intakes as follows:
Data availability
Intakec , n, a, s = intakec , n × scalarc , n, a, s The AFCD is open access and can be found at: https://dataverse.har-
vard.edu/dataverse/afcd. All other nutrient data were sourced from
We used the same process to disaggregate intakes for DHA+EPA and the USDA FoodData Central (https://fdc.nal.usda.gov/) or the GND
vitamin B12 but used the country-level and age-sex-level means derived as described in Methods. For sub-national data evaluation, data was
from SPADE habitual intakes described above. See Supplementary Table sourced from the following locations: FAO/GIFT: http://www.fao.org/
6 for details on crosswalking the Aglink–Cosimo and GENuS outputs. gift-individual-food-consumption/data-and-indicator/en/; NHANES:
We then used the SPADE habitual intake outputs to characterize https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx?
the distribution of nutrient intakes within each age–sex group. The BeginYear=2017; ENSANUT: https://ensanut.insp.mx/encuestas/ensa-
habitual intake data and associated statistical probability distributions nut2016/descargas.php; China Health and Nutrition Survey: https://
are incomplete across all country–nutrient–age–sex combinations www.cpc.unc.edu/projects/china/data/datasets; Uganda: https://doi.
(Supplementary Fig. 6) so we filled gaps by imputing data from the near- org/10.7910/DVN/FOYZBL; Burkina Faso: https://doi.org/10.7910/
est neighbour (37% of age–sex groups). First, we filled within-country DVN/5CXCLX. Proprietary input datasets protected by data-sharing
gaps by borrowing intake distributions, in order of preference, from agreements (that is, the GND) are not posted in these repositories. All
the nearest age group within a sex and country; the opposite sex from processed outputs and non-proprietary raw inputs are available on
within a country; and the nearest country geographically and/or socio- GitHub. The data associated with the diversity disaggregation is avail-
economically (Supplementary Fig. 7). We then mapped these to the able at https://github.com/cg0lden/Fisheries-Nutrition-Modeling. The
rest of the world, based on UN sub-regions, with a few expert-identified data associated with the SPADE analysis is available at https://github.
modifications (Supplementary Fig. 8). com/cg0lden/subnational_distributions_BFA. The data associated
with the health impacts analysis is available at https://github.com/
Health impact modelling approach alonshepon/Health-Benefit-Calculation-BFA.
SEVs integrate relative risks of sub-optimal diets with actual intake
distributions28. They estimate the population-level risk related to diets
and compare it to a population in which everyone is at a maximal risk Code availability
level, giving values ranging from 0% (no risk) to full population-level The code associated with the diversity disaggregation is available at
risk (100%). For DHA+EPA, we used the updated Institute for Health https://github.com/cg0lden/Fisheries-Nutrition-Modeling. The code
Metrics and Evaluation relative risk curves that are associated only associated with the SPADE analysis is available at https://github.com/
with ischaemic heart disease and have different values for adolescent cg0lden/subnational_distributions_BFA. The code associated with the
and adult subpopulations (with no risk for children). These relative health impacts analysis is available at https://github.com/alonshepon/
risk curves capture mild risk associated with consumption of omega-3 Health-Benefit-Calculation-BFA.
long-chain polyunsaturated fatty acids under 0.4 g per day28. For inad-
equate micronutrient intake risk assessment, we derived continuous 36. Pieralli, S., Pérez Domínguez, I., Elleby, C. & Chatzopoulos, T. Budgetary impacts of
adding agricultural risk management programmes to the CAP. J. Agric. Econ. 72, 370–387
relative risk curves for iron, zinc, calcium and vitamin A, based on the (2021).
probability approach for calculating inadequate intake, often a pre- 37. Ahern, M., Thilsted, S. H. & Oenema, S. The Role of Aquatic Foods in Sustainable Healthy
cursor to micronutrient deficiencies42. To evaluate the risk of inad- Diets. Discussion Paper (UN Nutrition, 2021).
38. Fluet-Chouinard, E., Funge-Smith, S. & McIntyre, P. B. Global hidden harvest of freshwater
equate micronutrient intake, distributions of intake are compared fish revealed by household surveys. Proc. Natl Acad. Sci. USA 115, 7623–7628 (2018).
against requirements. The latter is defined as a continuous risk curve 39. Dekkers, A. L., Verkaik-Kloosterman, J., van Rossum, C. T. & Ocké, M. C. SPADE, a new
statistical program to estimate habitual dietary intake from multiple food sources and
that has a value of 1 at low intakes, 0.5 at the relevant estimated aver-
dietary supplements. J. Nutr. 144, 2083–2091 (2014).
age requirement (EAR) and zero at large intakes. These absolute risk 40. Delignette-Muller, M. L. & Dutang, C. fitdistrplus: an R package for fitting distributions.
curves are based on the cumulative normal distribution function of J. Stat. Softw. 64, 1–34 (2015).
41. Smith, M. R., Micha, R., Golden, C. D., Mozaffarian, D. & Myers, S. S. Global Expanded
requirements43 with a mean at the EAR and a coefficient of variation of
Nutrient Supply (GENuS) Model: a new method for estimating the global dietary supply
10%. The latter value is used when more information on exact nutrient of nutrients. PLoS ONE 11, e0146976 (2016).
requirement is unavailable42,44. The prevalence of risk at the population 42. Institute of Medicine (US) Subcommittee on Interpretation and Uses of Dietary Reference
Intakes; Institute of Medicine (US) Standing Committee on the Scientific Evaluation of
level is derived by computing the expected micronutrient deficiency
Dietary Reference Intakes. DRI Dietary Reference Intakes: Applications in Dietary
across the entire population43, by applying an integral of the intake Assessment (National Academies Press, 2000).
distribution per age–sex–location–nutrient multiplied by its specific 43. Carriquiry, A. L. Assessing the prevalence of nutrient inadequacy. Public Health Nutr. 2,
23–34 (1999).
relative risk. The values derived range from 0 to 1, and evaluates the
44. Román-Viñas, B. et al. Overview of methods used to evaluate the adequacy of nutrient
risk of inadequate intake, as SEV, on a population level from no risk (0) intakes for individuals and populations. Br. J. Nutr. 101, S6–S11 (2009).
to maximal (1; everyone is at risk). Estimated average requirements 45. FAO/WHO. Vitamin and Mineral Requirements in Human Nutrition (World Health
Organization, 2004).
were derived from several sources45–47. Because zinc and iron require- 46. FAO/WHO. Guidelines on Food Fortification with Micronutrients (eds Allen, L., de Benoist,
ments depend on other dietary factors (for example, inhibitors such B., Dary, O. & Hurrell, R.) (World Health Organization, 2006).
as phytate), we used three levels for each nutrient, based on overall 47. Institute of Medicine. Dietary Reference Intakes: The Essential Guide to Nutrient Requirements
(eds Otten, J. J., Hellwig, J. P. & Meyers, L. D.) (National Academies Press, 2006).
diets, which crudely divide between diets based on their cereals and 48. Gibson, R. S., King, J. C. & Lowe, N. A review of dietary zinc recommendations. Food Nutr.
animal-source food intakes48,49. We then assigned each country to their Bull. 37, 443–460 (2016).
Article
49. Smith, M. R. & Myers, S. S. Impact of anthropogenic CO2 emissions on global human L. Jakobsen, C. Costello, S. Funge-Smith, A. Castellanos, C. Batis, J. Rivera, Y. Li, W. Willett,
nutrition. Nat. Clim. Change 8, 834–839 (2018). A. Dekkers, M. Moursi, H. Embke, A. Robertson, J. J. Hallo and K. Manning. Any use of trade,
50. Chua, H.-W., Wong, A., & Shek, Daniel T. L. In Encyclopedia of Quality of Life and firm, or product names is for descriptive purposes only and does not imply endorsement by
Well-Being Research (ed. Michalos, A. C.) 6047–6051 (Springer, 2014). the US Government.
51. Institute of Medicine. Dietary Reference Intakes for Thiamin, Riboflavin, Niacin, Vitamin B6,
Folate, Vitamin B12, Pantothenic Acid, Biotin, and Choline (National Academies Press, 1998). Author contributions C.D.G. and S.H.T. conceptualized the research idea, with substantial
52. Doets, E. L. et al. Systematic review on daily vitamin B12 losses and bioavailability for methodological and design input from J.Z.K., A.S., C.M.F., D.F.V. and H.M. Data acquisition and
deriving recommendations on vitamin B12 intake with the factorial approach. Ann. Nutr. compilation was conducted by subgroups for the Aquatic Foods Composition Database
Metab. 62, 311–322 (2013). (C.D.G., J.Z.K., C.D., H.K., K.J.F., M.K. and D.F.V.), Global Nutrient Database (H.M.), Aglink–Cosimo
model (H.M.), FAO Fish model (P.C., S.V. and M.B.), species disaggregation models (E.F.-C.,
E.A.N., J.A.G., A.J.L., D.F.V., J.G.E. and C.D.G.), sub-national distribution model (S.P., C.D.G., L.C.
Acknowledgements This paper is part of the Blue Food Assessment (BFA) (https://www.
and S.B.), and health impact models (A.S., C.D.G., G.D. and E.B.R.). The food systems modelling
bluefood.earth/), a comprehensive examination of the role of aquatic foods in building healthy,
was led by H.M. and P.C.; sub-national distributions modelling was led by S.P. and S.B.; and the
sustainable and equitable food systems. The BFA was supported financially by the Builders
health impact modelling was led by A.S., C.M.F. and G.D. C.D.G. drafted the original
Initiative, the MAVA Foundation, the Oak Foundation and the Walton Family Foundation, and
manuscript, and all co-authors edited and revised the writing.
has benefitted from the intellectual input of the wider group of scientists leading other
components of the BFA. We are also grateful for the financial support of the National Science
Foundation (CNH 1826668 to C.D.G., J.A.G. and J.G.E.) and the John and Katie Hansen Family Competing interests The authors declare no competing interests.
Foundation (C.D.G. and D.F.V.). Support was provided to SP and HK by a National Institutes of
Health Departmental Training Grant in Academic Nutrition (2T32DK007703-26). This work was Additional information
also undertaken as part of the CGIAR Research Program (CRP) on Fish Agri-Food Systems Supplementary information The online version contains supplementary material available at
(FISH), led by WorldFish and contributing to the WorldFish 2030 Research and Innovation https://doi.org/10.1038/s41586-021-03917-1.
Strategy: Aquatic Foods for Healthy People and Planet and the One CGIAR. This work is a Correspondence and requests for materials should be addressed to C.D.G.
product of the new Planetary Health laboratory at the Harvard T. H. Chan School of Public Peer review information Nature thanks Lotte Lauritzen, Max Nielsen and Albert Tacon for their
Health. For data sharing, visualization support and comments on study design, analysis and contribution to the peer review of this work.
early drafts of writing, we thank J. Schmidhuber, H. Lescinsky, A. Afshin, T. Beal, S. Pires, Reprints and permissions information is available at http://www.nature.com/reprints.
Extended Data Fig. 1 | Difference in daily per capita intake of various production scenarios, with and without fully accounting for species diversity.
nutrients from increasing aquatic animal-source food production and In the boxplots, the solid line indicates the median, the box indicates the
fully accounting for species diversity. The maps show the difference in mean interquartile range (IQR; 25th and 75th percentiles), the whiskers indicate 1.5
nutrient intakes in 2030 under the high and baseline production scenarios times the IQR, and the points beyond the whiskers indicate outliers. Countries
when fully accounting for species diversity. Values greater than zero indicate smaller than 25,000 km2 are illustrated as points (small European countries
higher nutrient intake under the high production scenario. Values less than excluded). All European Union (EU) member countries have the same value
zero indicate lower nutrient intake under the high production scenario. The because they are modelled as a single economic unit in the Aglink-Cosimo
boxplots show the difference in mean nutrient intakes in 2030 under both model (n=164 independent countries remain for comparison).
Article
Extended Data Fig. 2 | Difference in 2030 food consumption under the base and high production scenarios (part 1). Mean daily per capita food consumption
in 2030 under the (A) base and (B) high production scenarios and (C) the difference in consumption between the high production and base scenarios.
Extended Data Fig. 3 | Difference in 2030 food consumption under the base (C) the difference in consumption between the high production and base
and high production scenarios (part 2). Mean daily per capita food scenarios.
consumption in 2030 under the (A) base and (B) high production scenarios and
Article
Extended Data Fig. 4 | Difference in 2030 food consumption under the base and high production scenarios (part 3). Mean daily per capita food consumption
in 2030 under the (A) base and (B) high production scenarios and (C) the difference in consumption between the high production and base scenarios.
Extended Data Fig. 5 | Difference in 2030 nutrient intakes under the base when accounting for the full diversity of nutrient compositions in seafood
and high production scenarios accounting for the full diversity of nutrient under the (A) base and (B) high production scenarios and (C) the difference in
compositions in seafood. The mean daily per capita nutrient intake in 2030 intakes between the high production and base scenarios.
Article
Extended Data Fig. 6 | The relationship between the difference in 2030 indicates the average Estimated Average Requirement (EAR) for all age-sex
health outcomes under the high and base production scenarios and base groups. Countries falling below this line often have more room for health
scenario status. Each point represents a country where point color indicates improvements than countries falling above this line. Counter-clockwise from
the difference in national micronutrient deficiency averages between the the top-left, the quadrants of each plot indicate countries with mean 2030
scenarios (blue=reduced deficiencies; red=increased deficiencies) and point intakes in the base scenario that are: (1) higher than the mean EAR and higher
size indicates the scale of nutrient deficiencies in the base scenario (small=few than the high production scenario; (2) higher than the mean EAR but lower than
deficiencies; large=many deficiencies). The vertical line indicates zero the high production scenario; (3) lower than the mean EAR and lower than the
difference in nutrient intakes between the high and base scenarios; positive high production scenario; and (4) lower than the EAR but higher than the high
values indicate increased nutrient intake under the high production scenario production scenario.
and negative values indicate reduced intake. The dashed horizontal line
Extended Data Fig. 7 | Summary exposure values (SEVs) in the high diagonal line indicates the 1:1 line. Points below this line indicate
production scenario with and without the diversity disaggregation. country-age-sex groups with lower SEVs with the diversity disaggregation.
Summary exposure values (SEVs) for each country-age-sex group in the high Points above this line indicate country-age-sex groups with higher SEVs with
production scenario with and without the diversity disaggregation. The the diversity disaggregation.
nature research | reporting summary
Corresponding author(s): Christopher D. Golden
Last updated by author(s): 8/9/2021
Reporting Summary
Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Research policies, see our Editorial Policies and the Editorial Policy Checklist.
Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
The exact sample size (Ŷ) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
KŶůǇĐŽŵŵŽŶƚĞƐƚƐƐŚŽƵůĚďĞĚĞƐĐƌŝďĞĚƐŽůĞůǇďǇŶĂŵĞ͖ĚĞƐĐƌŝďĞŵŽƌĞĐŽŵƉůĞǆƚĞĐŚŶŝƋƵĞƐŝŶƚŚĞDĞƚŚŽĚƐƐĞĐƚŝŽŶ͘
A description of all covariates tested

A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)
For null hypothesis testing, the test statistic (e.g. &, ƚ, ƌ) with confidence intervals, effect sizes, degrees of freedom and W value noted
'ŝǀĞWǀĂůƵĞƐĂƐĞǆĂĐƚǀĂůƵĞƐǁŚĞŶĞǀĞƌƐƵŝƚĂďůĞ͘
For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's Ě, Pearson's ƌ), indicating how they were calculated
KƵƌǁĞďĐŽůůĞĐƚŝŽŶŽŶƐƚĂƚŝƐƚŝĐƐĨŽƌďŝŽůŽŐŝƐƚƐĐŽŶƚĂŝŶƐĂƌƚŝĐůĞƐŽŶŵĂŶǇŽĨƚŚĞƉŽŝŶƚƐĂďŽǀĞ͘
Software and code

Policy information about availability of computer code
Data collection No code on collection.
Data analysis The code associated with the diversity disaggregation is available in this Github repository: https://github.com/cg0lden/Fisheries-Nutrition-
Modeling
The code associated with the SPADE analysis is available in this Github repository: https://github.com/cg0lden/subnational_distributions_BFA
The code associated with the health impacts analysis is available in this Github repository: https://github.com/alonshepon/Health-Benefit-
Calculation-BFA
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and
reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.
Data
April 2020
Policy information about availability of data

All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A list of figures that have associated raw data
- A description of any restrictions on data availability
All processed outputs and non-proprietary raw inputs are available on Github.
The data associated with the diversity disaggregation is available in this Github repository: https://github.com/cg0lden/Fisheries-Nutrition-Modeling
1
The data associated with the SPADE analysis is available in this Github repository: https://github.com/cg0lden/subnational_distributions_BFA
The data associated with the health impacts analysis is available in this Github repository: https://github.com/alonshepon/Health-Benefit-Calculation-BFA
nature research | reporting summary

Proprietary input datasets protected by data-sharing agreements (i.e., the GND) are not posted in these repositories.
Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf
Behavioural & social sciences study design

All studies must disclose on these points even when the disclosure is negative.
Study description This is a quantitative food systems model that harnesses a scenario-based approach to evaluate the contribution of aquatic foods in
human diets and nutrition. The model is described fully in the methods section.
Research sample All countries with available data were included in our food system models. All countries with open-access, nationally representative
data were included in our subnational distributions models.
Sampling strategy All countries with available data were included in our food system models. All countries with open-access, nationally representative
data were included in our subnational distributions models. Therefore, the strategy did not include stratification or exclusion.
Data collection N/A
Timing N/A
Data exclusions N/A
Non-participation N/A
Randomization N/A
Reporting for specific materials, systems and methods

We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.
Materials & experimental systems Methods

n/a Involved in the study n/a Involved in the study
Antibodies ChIP-seq
Eukaryotic cell lines Flow cytometry
Palaeontology and archaeology MRI-based neuroimaging
Animals and other organisms
Human research participants
Clinical data
Dual use research of concern
April 2020
2
nature research | editorial policy checklist
Corresponding author(s): Christopher D. Golden
Last updated by author(s): 8/9/2021
Editorial Policy Checklist

This form is used to ensure compliance with Nature Research editorial policies related to research ethics and reproducibility. For further information,
please see our editorial policies site. All relevant questions on the form must be answered.
Competing interests
Policy information about competing interests
Competing interests declaration

In the interest of transparency and to help readers form their own judgements of potential bias, Nature Research journals require authors to declare any
competing financial and/or non-financial interest in relation to the work described in the submitted manuscript.
We declare that none of the authors have competing financial or non-financial interests as defined by Nature Research.
We declare that one or more of the authors have a competing interest as defined by Nature Research.
Authorship
Policy information about authorship
Prior to submission all listed authors must agree to all manuscript contents, the author list and its order and the author contribution statements. Any changes to
the author list after submission must be approved by all authors.
We have read the Nature Research Authorship Policy and confirm that this manuscript complies.
Data availability
Policy information about availability of data
Data availability statement

All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A list of figures that have associated raw data
- A description of any restrictions on data availability
We have provided a full data availability statement in the manuscript.
Mandated accession codes (where applicable)

Confirm that all relevant data are deposited into a public repository and that accession codes are provided.
All relevant accession codes are provided Accession codes will be available before publication No data with mandated deposition
Code availability
Policy information about availability of computer code
Code availability statement

For all studies using custom code or mathematical algorithm that is deemed central to the conclusions, the manuscript must include a statement under the
heading "Code availability" describing how readers can access the code, including any access restrictions. Code availability statements should be provided as a
separate section after the data availability statement but before the References.
We have provided a full code availability statement in the manuscript
July 2020
1
Data presentation
nature research | editorial policy checklist

For all data presented in a plot, chart or other visual representation confirm that:
n/a Confirmed
Individual data points are shown when possible, and always for Ŷ ч 10
The format shows data distribution clearly (e.g. dot plots, box-and-whisker plots)
Box-plot elements are defined (e.g. center line, median; box limits, upper and lower quartiles; whiskers, 1.5x interquartile range; points, outliers)
Clearly defined error bars are present and what they represent (SD, SE, CI) is noted
Image integrity
Policy information about image integrity
We have read Nature Research's image integrity policy and all images comply.
Unprocessed data must be provided upon request. Please double-check figure assembly to ensure that all panels are accurate (e.g. all labels are correct, no
inadvertent duplications have occurred during preparation, etc.).
Where blots and gels are presented, please take particular care to ensure that lanes have not been spliced together, that loading controls are run on the same
blot, and that unprocessed scans match the corresponding figures.
Additional policy considerations

Some types of research require Does not apply Involved in the study
additional policy disclosures. Please
indicate whether each of these Macromolecular structural data
apply to your study. If you are not Unique biological materials
certain, please read the
Research animals and/or animal-derived materials that require ethical approval
appropriate section before
selecting a response. Human embryos, gametes and/or stem cells
Human research participants
Clinical data
I certify that all the above information is complete and correct.

Typed signature Christopher D Golden Date 8/9/2021
July 2020
This checklist template is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium
or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images
or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in
the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the 2
copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Aquatic Foods To Nourish Nations: Article

Uploaded by

Copyright:

Available Formats

Aquatic Foods To Nourish Nations: Article

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Aquatic Foods To Nourish Nations: Article

Uploaded by

Copyright:

Available Formats

Article

Aquatic foods to nourish nations

Nature | Vol 598 | 14 October 2021 | 315

(water spinach; Ipomoea aquatica), algae (seaweed) and other aquatic

and geographic regions. Here we focus on AASFs, which constitute the

of nutrients, including minerals (for example, calcium, iron and zinc),

Aquatic foods to benefit human health Omega-3

AASFs improve human health through at least three pathways: by

316 | Nature | Vol 598 | 14 October 2021

c Poultry consumption d Egg consumption

e Dairy consumption f Non-aquatic animal-source food consumption

Nature | Vol 598 | 14 October 2021 | 317

ΔSEVs (%) ΔSEVs (%)

ΔSEVs (%) ΔSEVs (%)

ΔSEVs (%) ΔSEVs (%)

Women and girls Men and boys ΔMillions of people

318 | Nature | Vol 598 | 14 October 2021

Nature | Vol 598 | 14 October 2021 | 319

320 | Nature | Vol 598 | 14 October 2021

A description of all covariates tested

Software and code

Policy information about availability of data

nature research | reporting summary

Behavioural & social sciences study design

Data collection N/A

Data exclusions N/A

Reporting for specific materials, systems and methods

Materials & experimental systems Methods

Editorial Policy Checklist

Competing interests declaration

Data availability statement

Mandated accession codes (where applicable)

Code availability statement

nature research | editorial policy checklist

Additional policy considerations

I certify that all the above information is complete and correct.

You might also like