Vulnerability of Blue Foods To Human-Induced Environmental Change

nature sustainability
Article https://doi.org/10.1038/s41893-023-01156-y
Vulnerability of blue foods to

human-induced environmental change
Received: 25 January 2022 Ling Cao 1,17 , Benjamin S. Halpern 2,3,17, Max Troell 4,5,17, Rebecca Short4,17,
Cong Zeng 6,17, Ziyu Jiang 6,17, Yue Liu 6,17, Chengxuan Zou6, Chunyu Liu 6,
Accepted: 24 May 2023
Shurong Liu6, Xiangwei Liu 6, William W. L. Cheung 7, Richard S. Cottrell 8,
Published online: 26 June 2023 Fabrice DeClerck 4,9, Stefan Gelcich 10, Jessica A. Gephart 11,
Dakoury Godo-Solo11, Jessie Ihilani Kaull12, Fiorenza Micheli 13,
Check for updates
Rosamond L. Naylor 14, Hanna J. Payne 15, Elizabeth R. Selig 15,
U. Rashid Sumaila 16 & Michelle Tigchelaar 15
Global aquatic or ‘blue’ foods, essential to over 3.2 billion people, face
challenges of maintaining supply in a changing environment while adhering
to safety and sustainability standards. Despite the growing concerns
over their environmental impacts, limited attention has been paid to how
blue food production is influenced by anthropogenic environmental
changes. Here we assess the vulnerability of global blue food systems to
predominant environmental disturbances and predict the spatial impacts.
Over 90% of global blue food production faces substantial risks from
environmental change, with the major producers in Asia and the United
States facing the greatest threats. Capture fisheries generally demonstrate
higher vulnerability than aquaculture in marine environments, while the
opposite is true in freshwater environments. While threats to production
quantity are widespread across marine and inland systems, food safety
risks are concentrated within a few countries. Identifying and supporting
mitigation and adaptation measures in response to environmental stressors
is particularly important in developing countries in Asia, Latin America and
Africa where risks are high and national response capacities are low. These
findings lay groundwork for future work to map environmental threats
and opportunities, aiding strategic planning and policy development for
resilient and sustainable blue food production under changing conditions.
Blue foods, defined as fish, shellfish, plants and algae from fishing affects the environment and resource systems, from local to global
and farming in marine and freshwater ecosystems, play an important scales7–12. Historically, blue foods have been heavily underrepresented
role in food and nutrition security for billions of people while sup- in global food systems, models and assessments1,4,13,14. Such oversights
porting livelihoods, economies and cultures around the world1–6. As a can lead to policies that are based largely on terrestrial foods, missing
key source of animal proteins, essential fatty acids and critical micro- the opportunity to include blue foods in efforts to build a more diverse
nutrients, blue foods make vital contributions to averting the triple and resilient food system. In a world facing increasingly frequent and
burden of malnutrition, particularly in low-income and Small Island severe disturbances, blue foods play an increasingly important role
Developing States (SIDS)3. Central to scientific and policy discussions in avoiding human food crises that terrestrial systems alone may not
on the sustainability of food production has been how this production be able to address. Overlooking blue foods in food policy discussions
A full list of affiliations appears at the end of the paper. e-mail: caoling@xmu.edu.cn
Nature Sustainability | Volume 6 | October 2023 | 1186–1198 1186

also increases the chances of negative and suboptimal environmental benthic and artisanal rivers, and eight were for aquaculture, such as
outcomes that could arise from ignoring the many links between ter- cage finfish and pond crabs. Our goal in delineating these archetypes
restrial and aquatic systems15. was to represent the predominant aquatic production systems and
Given the growing emphasis on the role of blue foods in the current account for broad differences in their responses to stressors, including
and future transitions to sustainable food systems, inadequate atten- production links to specific biomes or habitats, the degree of produc-
tion has been paid to the vulnerability of blue food systems to envi- tion intensity and functional traits of the species produced.
ronmental disturbances, with the possible exception of assessments We then performed a literature review using the Google Scholar
focused on the current and projected impacts of climate change16–19. keyword searching tool to identify the dominant anthropogenic stress-
Blue food production is tightly coupled with environment and resource ors that have direct impacts on the quantity or the quality (safety) of
systems. Wild fish stocks targeted by capture fisheries depend on blue foods at the production stage of the value chain (see ‘Stressor
marine, coastal and inland freshwater ecosystems, while most aqua- identification’ in Methods and Supplementary Data 1). Ten key stress-
culture also relies on healthy aquatic and terrestrial environments ors affecting the production quantity of blue foods were identified,
to provide suitable farming environments as well as fish feed, seed including five climatic stressors (warming, acidification, sea level rise,
(broodstocks and larvae) and other supporting functions to ensure severe weather events and altered precipitation) and five non-climatic
survival and production4,11,18,20–22. These close ties to the environment stressors (hypoxia, eutrophication, diseases, invasion and parasites). In
suggest that blue food production may be particularly susceptible addition, seven stressors were identified as primary concerns for blue
to human-induced environmental change, from changes in habitat, food safety, including three biological stressors (harmful algal bloom
alterations to water quality and quantity, and pollution that affects toxins, non-indigenous bacteria introduced by humans into aquatic
growth, among other factors. systems and indigenous bacteria influenced by anthropogenic drivers
Blue foods are highly diverse, encompassing more than 540 spe- such as climate change) and four chemical stressors (heavy metals, in
cies farmed or harvested in freshwater and over 2,190 species pro- this context referring to mercury; persistent organic pollutants (POPs);
duced in marine systems2–4. This diversity provides both potential pesticides; and antibiotics). Stressors were also defined as either press
environmental liabilities and advantages. Notably, the specific effects (persistent and long-term, as with sea level rise (SLR)) or pulse (brief
of anthropogenic pressures on blue foods probably vary widely across and short-term although potentially frequent, as with severe weather
species and systems, with unique sensitivities to different combi- events)27 (Supplementary Fig. 1). Harmful algal bloom (HAB) toxins
nations of stressors and subsequent direct and indirect effects; the are unique among this set in having a dual effect on both the quantity
liability then is that key foods may be particularly vulnerable to anthro- and quality of blue foods28. Other stressors not included here, such as
pogenic pressures, but the advantage may lie in greater sustainability different types of pathogens and a number of chemical pollutants, may
and resilience through portfolio effects11,23. be of localized importance but are not globally prevalent or have been
Here we focus on how environmental disturbances, characterized identified as severely lacking evidence.
by 17 anthropogenic stressors, are expected to affect the quantity Once key archetypes and stressors were defined, a multifactor
(volume produced) and quality (food safety) of blue foods produced spatial analysis25,26 was conducted to quantify the spatial exposure
in the wild and through farming. Following a vulnerability assessment of blue food production to these stressors, including presence and
framework24 that defines ‘vulnerability’ as a combination of sensitivity intensity29. The exposure level for each stressor showed notable spatial
(that is, susceptibility to external stresses) and adaptive capacity (that variability. Some stressors exhibited localized concentrations of expo-
is, the intrinsic ability to adjust to potential damage, to take advantage sure intensity (for example, acidification, hypoxia, parasites and inland
of opportunities, or to respond to consequences of environmental warming in Europe, and inland precipitation variability in Asia), while
variability and external disturbance), and ‘impact’ as a function of others were more dispersed (Fig. 1). Thailand was found to be highly
vulnerability and exposure (that is, external stresses), we rely on expert exposed to pathogens associated with blue food production. Cyprus
elicitation, literature review and spatial analysis to investigate the spa- suffers from high exposure to antibiotics. Trinidad and Tobago has the
tial impacts of environmental change on global blue food production highest exposure to mercury across freshwater and marine water bod-
systems (Methods). We first use literature review and expert judgement ies, mainly attributed to the widespread small-scale gold mining in Latin
to identify the most relevant and well-documented anthropogenic America30. Belgium and the Netherlands are particularly exposed to
stressors and corresponding geospatial indicators. Next, we estimate harmful parasites. Climate-linked stressors on the production quantity
exposure scores for each production system–stressor combination at demonstrate high exposure intensity across a wide geographic range,
the national level using a multifactor spatial analysis25,26 and assess the particularly in the Northern Hemisphere, while non-climatic stressors
vulnerability of blue foods to environmental change. Finally, we com- show similarly high intensity across just a few countries (Supplemen-
pute an impact index by country to map how environmental stresses tary Fig. 2). For food safety, most countries show moderate to high
impact the quantity and safety of blue food production systems. Our exposure to inland chemical stressors and low to moderate exposure
analysis fills a critical need to improve our understanding of the poten- to biological stressors and marine chemical stressors, with only a few
tial feedback between environmental change and blue food systems, countries in Southern Europe and Africa showing the highest intensity
and helps identify priority areas for research and action. In addition, of exposure (Supplementary Fig. 2). Our indicator data for environmen-
we identify key opportunities and challenges for reducing impacts tal stressors were derived from historical observations from different
and mitigating risk to ensure the continued provision of blue foods public recordings and may contain observational bias. Some of the
in the future. unique exposure levels were, however, supported by the literature;
for example, severe exposure to antibiotics in Cyprus primarily arises
Exposure of blue foods to environmental stresses from the high consumption of anti-infective agents31.
Given their great diversity, blue foods may experience exposure to
anthropogenic environmental stresses in different ways. We first used Vulnerability of blue foods
expert assessment workshops with paper co-authors to classify blue We assessed the likely vulnerability of each blue food archetype to each
food production into groups according to the environment (marine or stressor using targeted literature reviews and drawing on expert assess-
freshwater), production method (capture or aquaculture) and produc- ment to score each stressor–archetype interaction as high, medium,
tion system (for example, pond, cage), leading to 15 production arche- low or none (see ‘Vulnerability’ in Methods and Supplementary Data
types (see ‘Blue food system and archetype classification’ in Methods). 2). The purpose of the vulnerability assessment was not to describe
Seven of these were for capture fisheries, such as marine commercial how each archetype is currently functioning, but rather to focus on the

a Greenland
b
Cyprus
Acidification
Poland
Antibiotics
Italy
Sweden
Russia Spain
Denmark Tunisia
Belgium
Thailand
Mexico
Diseases
Poland Cyprus
China
Antibiotics
Italy
Japan
Tunisia
USA Belgium
France
Eutroph-
Algeria
ication
UK
Sweden
Togo
Lithuania
Albania
Bacteria
Norway Benin
Belgium
Eutroph-
Montenegro
ication
Slovenia
Lebanon Bahamas
Israel
Benin
Denmark Albania
Bacteria
Germany
Hypoxia
Norway Togo
Slovenia Montenegro
Belgium Bosnia and Herz.
USA
Australia Polynesia
Invasion
France Philippines
Japan
HAB
France
Canada
USA
Netherlands Canada
Belgium
Parasites
UK Trinidad and Tob.

Germany
Denmark Bangladesh
Mercury
Bhutan
Belgium
Equatorial Guinea
Netherlands
Parasites
Germany Indonesia
Denmark
Norway Rep. of the Congo
Grenada
Mercury
Papua New Guinea
Precipitation
Cambodia Trinidad and Tob.

Vietnam Cameroon
United Arab Emirates Nigeria
Djibouti
China Lebanon
India S.Korea
weather
Pesticide
Severe
USA Ukraine
Australia
Indonesia Czech Rep.
Bosnia and Herz.
Japan
Australia
weather
Bosnia and Herz.

Severe
Philippines
Mexico Poland Pesticide
China Belgium
Bangladesh
Solomon Islands
Estonia
Sea level
Micronesia
Finland
rise
Estonia Eswatini
Palau Albania
POPs
Montenegro
Russia
Lesotho
Warming
Greenland
Belarus Greece
Ukraine
Kuwait
Montenegro
Russia Belgium
POPs
Warming
Canada Albania
Greenland
Togo
Iraq
Kuwait Ghana
0 0.25 0.50 0.75 1.00 0 0.25 0.50 0.75 1.00

Standardized exposure Standardized exposure
intensity Marine Freshwater Both intensity
Fig. 1 | Top five countries with the highest exposure to anthropogenic stresses. a,b, Exposures are in terms of production quantity (a) and food safety (b).
Scores were standardized and should only be compared within stressors.

FW capture FW culture Marine capture Mariculture
hic
ic
ves
ds
elag
h
wns
es
gae
ent
nds
infis
ish
fish
ival
tlan
bs
xed
ak
ers
es
m. b
m. p
finf
pra
cra
o
m. l
an b
e fin
an a
an f
. we
. lak
h. p
. riv
. mi
d
d
Com
Com
Com
Oc e
Oc e
Oc e
Cag
Pon
Pon
Pon
Arti
Arti
Arti
Arti
Exc
Diseases
Parasites
Non-climatic
Invasion
Production quantity
Eutrophication
Hypoxia
Altered precipitation
Warming
Climatic
Acidification
Sea level rise
Severe weather events
Mercury
Chemical
POPs
Antibiotics
Food safety
Pesticide
HAB toxins
Biological
Non-indigenous bacteria
Indigenous bacteria
Vulnerability High Medium Low None Confidence Sufficient Partial Poor None
Fig. 2 | Vulnerability scores of 15 blue food archetypes to anthropogenic stressors affecting production quantity and food safety. Confidence scores are
provided for each archetype–stressor assessment on the basis of the availability of evidence determined through the literature review and expert input.
FW, freshwater; Arti., artisanal; Comm., commercial; Exch., exchange.
potential reduced functioning of each archetype under a given stress. warming, as frequently targeted fish communities in lakes and coral
Archetypes were vulnerable to different combinations of stressors and reefs are more sensitive to warming34,35. In general, marine fisheries
the extent of vulnerability varied (Fig. 2). Capture fisheries generally were more vulnerable to climatic stresses particularly warming and
demonstrated higher vulnerability than aquaculture. Of all archetypes, acidification, while aquaculture was more vulnerable to non-climatic
marine fisheries, particularly artisanal mixed and commercial benthic stresses such as diseases and hypoxia.
fisheries, were the most vulnerable, mainly due to mercury, acidifica- Vulnerability scores were generally lower across archetypes for
tion and warming. Seaweed mariculture, which accounts for over 99.5% food safety stressors than for production quantity. However, given
of global algae production, was the least vulnerable, demonstrating its the potential human health consequences of reduced food safety,
outstanding potential for mitigation and adaptation to environmental even low-vulnerability blue food groups should be of concern in food
disturbances. system management4. Mercury contamination was the most pervasive
In terms of production quantity, freshwater production was gener- stressor creating the highest vulnerabilities across archetypes. Impacts
ally more vulnerable than marine production. Freshwater prawn ponds of mercury were exceptionally high for marine commercial benthic
and marine exchange ponds appeared most vulnerable due to disease and pelagic fisheries, resulting in them being the most vulnerable
and hypoxia stresses. Freshwater fisheries were more susceptible to archetypes in terms of food safety, partially due to the trophic lev-
the introduction of invasive species and altered precipitation than els targeted. The growing list of chemicals (for example, antibiotics)
others, which was consistent with the literature32,33. Indeed, altered classed as persistent organic pollutants (compounds that persist in the
precipitation was the only climatic stressor that primarily contrib- environment, being resistant to degradation) and the prevalence of
uted to the vulnerability of the freshwater archetypes, with riverine non-indigenous bacteria were of broad concern across the archetypes.
and wetland fisheries being the most vulnerable. Across all stressors, Vulnerability to pesticide use was highest in freshwater and brackish
hypoxia demonstrated the broadest impacts with associated effects of systems. Anthropogenically linked increases in indigenous bacteria
low water flows, increased water temperature and increased nutrient were largely a marine issue of lower concern than non-indigenous intro-
pollution. Artisanal fisheries were most vulnerable to surface water ductions, and benthic fisheries demonstrated the highest vulnerability.

Capture fisheries, particularly freshwater ones, showed higher vulner- while the low impact group was dominated by SIDS. Both freshwa-
ability to food safety stressors than aquaculture. The confidence scores ter production quantity and safety, particularly from aquaculture,
indicated a lower level of evidence related to food safety stressors at were moderately to highly affected in all country clusters except SIDS,
the level of archetypes than for production quantity, which was largely mainly attributed to warming, altered precipitation and eutrophica-
due to the reliance on monitoring efforts at the environment rather tion. Meanwhile, in SIDS, marine production quantity was most affected
than the production system or species level. due to high exposure to sea level rise. Our results highlight the need to
This assessment took a holistic view of the vulnerability of blue design effective management systems to enhance adaptation to climate
food archetypes where relevant information was available, but change and to support climate-resilient fisheries and aquaculture for all
trade-offs were necessary for data compilation on food safety stress- countries. In addition, more stringent control of inland eutrophication
ors including pesticides, antibiotics and HAB toxins. Given that in will be priorities for all countries except SIDS.
many cases the specific food safety stressor–archetype interaction is a To identify and prioritize adaptation and resilience-building
relatively niche piece of information to access, we needed to generalize strategies, particularly for countries with high estimated impacts, we
from more specific foci for the vulnerability assessment. The charac- assessed national response capacities reflecting their social, political
teristics of some food safety stressors (for example, mercury, POPs and and economic potential or ability to cope and adapt to environmental
bacteria) demonstrating substantial, fine-scale spatial and temporal stresses (Supplementary Fig. 5). We then explored the relationship
fluctuations, potentially exceeding thresholds for human health, could between estimated impacts on blue food production and national
not be sufficiently covered. Such details are utterly important when response capacity at the country level and prioritized outliers for miti-
considering vulnerability. In these cases, it was necessary to general- gation (Fig. 5). Regarding production quantity, Denmark, Belarus and
ize risk considerations. For example, mercury demonstrates global the United States appeared as mean-shift outliers with high response
airborne deposition (delivery of substances by air) in aquatic environ- capacity but disproportionately large risks. In terms of food safety,
ments but also notable point source (single identifiable source) inputs, Albania, Montenegro, and Bosnia and Herzegovina were identified as
which may impact vulnerability as well as exposure in very localized outliers due to extremely high impacts. Countries with the highest esti-
areas, such as in artisanal gold mining areas for freshwater systems36. mated impact but lowest national response capacity (that is, countries
in the upper left corner of Fig. 5) may be priorities for impact mitigation
Estimated impacts of anthropogenic stresses on and capacity building in terms of improved governance, economy and
blue foods social development to enable more resilient blue food production. In
Variations in blue food exposure and vulnerability generated large terms of production quantity, high impact and low-capacity countries
spatial differences in estimated total impact (as a function of exposure were found only in freshwater aquaculture. These countries, all of which
and vulnerability and determined by spatial analysis; see ‘Estimated are developing countries, including Uganda, Bangladesh, Eswatini,
impact’ in Methods) and key sources of impact across environments and Honduras and Guatemala, were mostly affected by altered precipitation
nations. Production quantity of aquaculture was predicted to be more and parasites. Therefore, coping with impacts of climate change and
severely impacted than capture fisheries in freshwater environments, specific harmful parasites in these countries should be prioritized to
while the opposite was true in marine environments (Fig. 3a). With at promote resilient aquaculture. With low national response capacities,
least one of the four production systems having standardized impact Togo and Benin face high food safety threats for freshwater aquaculture
scores above 0.8, blue food production quantity in leading producing and marine fisheries due to high exposure to inland pesticides, marine
countries such as China, Norway and the United States was estimated POPs and bacteria in both freshwater and marine water bodies. These
to be most affected. Across these countries, high exposure to inland acute impacts call for urgent and effective countermeasures to reduce
eutrophication, inland warming, altered precipitation and sea level and manage organic pollutants, control pesticide use and better screen
rise was mostly common. for food safety at production sites.
Whereas impacts on quantity were concentrated, impacts on
food safety were sporadically distributed across the globe, with rela- Improving future blue food vulnerability
tively consistent patterns between freshwater and marine production estimates
(Fig. 3b). Blue food safety in Albania and Montenegro was estimated Estimating the vulnerability of blue foods to anthropogenic stressors at
to be most impacted at the production stage, as more than three of the global scale necessarily faces data and methodological limitations.
their four production systems had standardized impact scores above Global datasets for selected key stressors are limited, constraining our
0.8. Impacts of POPs, bacteria and inland pesticides were particularly ability to assess exposure to certain stressors and most stressors at
acute in these two countries. Although a comparable distribution was subnational resolution. Furthermore, most vulnerability assessments,
found in production quantity impacts across continents, Oceania was including our study, remain primarily based on historical trends or
markedly lower than other regions in terms of risks to food safety, as it snapshots, which may lead to inadequate conclusions about future
has fewer potentially highly polluting industries in SIDS (Supplemen- risk37,38. Finally, data reported at different scales had to be resolved to
tary Fig. 3). A deeper dive into the impact profiles revealed that there the coarsest scale, resulting in some information reduction. For exam-
was no clear relationship between the impacts on quantity and those ple, indicators for climatic stressors are often available at finer spatial
on safety (Supplementary Fig. 4). Thus, the culmination of different resolution than for biological and chemical stressors; to address this, we
threats and the realized impacts of stressors on blue foods require aggregated high-resolution spatial data to calculate a national average
highly contextualized mitigation and adaptation strategies to ensure over space for the geospatial mapping, limiting our ability to address
stable production under environmental change. subnational variation. Despite data limitations, the uncertainty analysis
To identify particularly vulnerable groups of countries at risk from showed that our overall results were robust (Supplementary Fig. 6).
multiple forms of environmental change, we performed a cluster analy- A critical gap influencing the assessment of food safety risks is
sis of the overall impact for 222 countries and territories that reported accounting for the relationship between point source pollution, its dis-
blue food production over the past decade. We found four groups of tribution, and subsequent uptake and bioaccumulation in blue foods.
countries characterized by high, medium-high, medium-low and low Local-scale models exist for the distribution, accumulation and circu-
impacts (Fig. 4). The high and medium-high impact groups included lation of pollutants within ecosystems, but we lack regional or global
many of the major blue food producing countries on our planet (for scale high-resolution models that allow for the global assessment and
example, China, India, Vietnam, Norway and the United States). The comparisons done here. Building off the structure of existing datasets
medium-low impact group was dominated by landlocked countries, for chlorophyll, silicate, nutrients and sea surface temperature, such

Standardized impact score
0.0–0.2 0.2–0.4 0.4–0.6 0.6–0.8 0.8–1.0

a
Impact on quantity
FW capture FW culture
Marine capture Mariculture
b
Impact on safety
FW capture FW culture
Fig. 3 | Estimated impacts of anthropogenic stresses on global blue food production. a, Impacts on quantity. b, Impacts on safety. Results were standardized and
presented at the national level for each blue food production system. The four production systems included freshwater capture fisheries, freshwater aquaculture,
marine capture fisheries and mariculture.
as Bio-ORACLE39,40, to guide and standardize global data on pollutants impacts of stressors on blue foods. We assessed vulnerability only at
would substantially advance analytical options. Additionally, we need the production stage; however, future work needs to improve under-
continued research on the uptake, processing and storage of pollut- standing of post-production human health risk profiles through the
ants by aquatic organisms used for food throughout their life stages value chain and the pathways by which they accumulate into organisms
to better assess the vulnerabilities of blue foods and design mitigation before human consumption.
strategies. A critical trade-off exists for organisms that are vulnerable
to human health risks, such as stationary filter-feeding bivalves, but Policy implications
that may also provide key nutrition3 through high nutrient density or Blue foods are an indispensable part of sustainable and equitable food
accessibility to malnourished populations, and may be a sustainable systems and healthy diets3,4, making it important to understand the
alternative compared with production of other organisms7. In such extent to which anthropogenic environmental stressors may impede
cases, localized capacity for mitigation of such risks will be a key deter- production of and access to adequate and safe seafood. This paper
mining factor in their promotion. Combined, the above information presents a comprehensive global analysis based on a composite vulner-
will add to expanding understanding of sustainability and nutritional ability index to illustrate how pervasive anthropogenic stressors may
contributions for better planning of what species to harvest, avoid or affect the quantity and safety of fisheries and aquaculture production.
mitigate, as well as developing strategies for when and where to harvest Our analysis suggests that in marine environments, capture fisheries
or cultivate them. are generally more vulnerable than aquaculture, while the opposite is
Additional challenges arise in accounting for the synergistic or true in freshwater environments. Highly vulnerable production systems
cumulative impacts from multiple stressors that often vary among exist across all continents. Potential effects of environmental change
stressor combinations and across temporal and spatial scales38,41. These on production quantity and safety differ in that production quantity
interactions remain poorly understood and need to be addressed in faces high threats in both marine and freshwater systems across many
future vulnerability assessments, particularly at the regional or local locations, whereas food safety receives elevated impacts across a nar-
level. Our results here represent conservative estimates of realized rower geographical focus. In both cases, there is a need to address

a
0.7
High impact
0.6
Standardized impact score

0.5
Medium-high impact
0.4
Medium-low impact
(landlocked) 0.3
0.2
0.1
Low impact
(SIDS)
0
y
ty
ty
ty
ty
tit
tit
tit
tit
fe
fe
fe
fe
an
an
an
an
sa
sa
sa
sa
qu
qu
qu
qu
e
re
re
ur
ur
ltu
tu
e
re
re
pt
pt
ur
ur
ul
ltu
tu
cu
ca
ca
pt
pt
ic
ul
cu
ca
ca
ar
FW
ic
FW
e
in
M
ar
FW
FW
ar
in
M
ar
M
b FW capture quantity
1.0
Mariculture safety FW culture quantity
0.5
Marine capture safety 0 FW capture safety
Mariculture quantity FW culture safety
Marine capture quantity
Medium low impact Low impact

High impact Medium high impact
(landlocked) (SIDS)
Fig. 4 | Cluster analysis based on the overall impact of all blue food systems high impact (n = 91 countries), medium-low impact (n = 39 countries) and low
for a total of 222 countries and territories. a,b, Heat map (a) and radar plot (b) impact (n = 59 countries). The list of countries of the four clusters is provided in
of standardized impact scores of the four clusters obtained by k-means analysis. Supplementary Data 3.
The four clusters are characterized by high impact (n = 33 countries), medium-
stressors regionally, as stressors cross boundaries, and strengthen be paid to developing countries in Asia (for example, Bangladesh),
national response capacity where most necessary depending on the Africa (for example, Togo) and Latin America (for example, Honduras)
level of dependence on marine versus inland production. where risks are high and national response capacities are low. These
Our analysis of blue food vulnerability at the national level can countries are more likely to suffer from environmental disturbances
help us understand how each nation suffers from different combina- and should prioritize both impact mitigation and capacity building
tions of stressors and its varying extent of vulnerability. Countries in terms of improved governance, economy and social development
identified as highly vulnerable include the major blue food producers to enable more resilient blue food production. In these countries,
in Asia (for example, China, Japan, India and Vietnam) that account for diversification of production (for example, expanding aquaculture
over 45% of global landings and 85% of global aquaculture production, in regions where it is still nascent and has low estimated impact) will
and thus require priority adaptation and mitigation actions to reduce be necessary unless sufficient mitigation and adaptation strategies
vulnerability and exposure to hazards. Particular attention should are implemented.

a
Thailand
1.00 FW capture 1.00
FW culture
USA
China
China
0.75 0.75
0.50 0.50
0.25 0.25
Predicted impact on quantity
Samoa
Singapore
Romania
0 Micronesia 0
0 0.25 0.50 0.75 1.00 0 0.25 0.50 0.75 1.00
1.00 Denmark 1.00

USA Norway Norway
0.75 Japan 0.75
0.50 0.50
0.25 0.25
Romania
0 0
0 0.25 0.50 0.75 1.00 0 0.25 0.50 0.75 1.00
1.00 Albania 1.00

FW capture FW culture Albania
0.75 0.75
Bangladesh
0.50 0.50
0.25 0.25
Predicted impact on safety
0 0
0 0.25 0.50 0.75 1.00 0 0.25 0.50 0.75 1.00
Bosnia and Herz.

1.00 Marine capture 1.00
Bosnia and Herz.
Mariculture
Benin Albania
Montenegro
0.75 0.75
0.50 0.50
Norway
0.25 0.25 Norway
Iceland
0 0
0 0.25 0.50 0.75 1.00 0 0.25 0.50 0.75 1.00
National response capacity
Production 104 106 108 Continent Africa America Asia Europe Oceania
Fig. 5 | Relationship between the estimated impact and national response see ‘Bonferroni outlier test’ in Methods). Dashed lines indicate the one-third
capacity of each blue food nation. a, Quantity. b, Safety. Circle size indicates division lines of the maximum values of the horizontal and vertical axes. Grey
the average blue food production (metric tons) of each nation for the past shadings along the axes indicate density curves (the total area under the density
10 yr. Circles with solid line borders and corresponding country names in bold curve represents probability). Most country names have been hidden to improve
indicate mean-shift outliers (extreme observations furthest from the mean value readability, and a detailed list of country rankings for each production system is
of the neighbourhood) based on Bonferroni outlier tests (P < 0.05, two-sided; provided in Supplementary Data 4.

Our findings provide key insights needed for decision-makers interactive workshops and online document sharing. During the assess-
to formulate strategic actions in their jurisdictions, particularly with ment, experts were first asked to answer the pre-designed questions
respect to specific vulnerabilities of blue foods disaggregated by independently. Facilitators then synthesized initial responses and
region and production archetype. For example, the prevalent shrimp returned anonymous results to all experts for discussion, allowing
aquaculture in Thailand faces substantial climate and disease risks, them to give a second round of feedback accordingly. Then the mode
and requires not only technological innovation in farming but also of the second round of responses was aggregated and shared with all
risk-sharing strategies such as price contracts and insurance42. These experts for final review and approval. The last step was repeated until
results also highlight the need for transboundary collaboration in none of the experts disagreed, thus reaching a consensus. Full details
mitigating stressors, particularly for food safety, given that risks can of the protocol are provided in the Supplementary Information.
be transmitted through interconnected ecosystems (across national
boundaries) and ultimately via the supply chain in a highly globalized Blue food system and archetype classification
food system that ‘redistributes’ our impact results to ultimate locations To represent the diversity of blue food systems and capture how each
where blue foods are consumed. Such transboundary issues highlight system will be affected by environmental stressors, we ran a first expert
the importance of international policy in maintaining sustainable blue assessment and asked experts to individually answer one question:
food production under anthropogenic environmental change. “What are the main blue food systems and archetypes?” We initially
Across all blue food archetypes, greater stakeholder engage- classified production into freshwater and marine systems, inclusive of
ment in efforts to understand, monitor and mitigate pressures on both capture fisheries and aquaculture. Archetypes were then devel-
different food types will be essential to bring higher-resolution infor- oped through a series of workshops with the co-author team, alongside
mation to bear on our understanding of vulnerability and exposure. discussions on the scope of the paper. Following such discussions,
Greater stakeholder engagement will also be important in efforts to it was agreed that categorization should aim to represent dominant
design effective management actions and enhance local adaptation global production systems and, in particular, to categorize these sys-
responses. Particularly for artisanal fisheries archetypes and heavily tems by aspects that affect their vulnerability to stressors. The defining
marine fisheries-dependent countries such as SIDS that often have characteristics of such systems are linked broadly to the habitats or
limited data but require urgent actions to cope with climate change43,44, environments utilized and the species targeted or cultured. Further
incorporation of local and indigenous knowledge will be critical inputs discussion refined these aspects to be broadly dominated by operation
for strategic planning and policy development. Indeed, this informa- type and target environment for fisheries, and operation type and taxa
tion is paramount for the one-third to one-half of the world that lies produced for aquaculture. A resulting initial long list of 28 production
within indigenous lands and territorial waters45. types was then refined for feasibility to the final 15 archetypes, with
Despite progress in developing adaptation strategies in blue food the aim to best capture differences with respect to how stressors may
systems to climate change19, substantial challenges remain in develop- impact production.
ing similar strategies for the complete set of stressors associated with The 15 archetypes are: freshwater capture fisheries (artisanal lakes,
anthropogenic environmental change. Our work provides an important artisanal rivers, artisanal wetlands and commercial lakes); freshwater
foundation to guide future blue food research to develop such strate- aquaculture (cage finfish, pond crabs, pond finfish and pond prawns);
gies and assess national-scale impacts to help inform context-specific marine capture fisheries (artisanal mixed, commercial benthic and
mitigation strategies. Translating this global work into regionally and commercial pelagic); and mariculture (ocean algae, ocean bivalves,
locally appropriate actions remains a key next step. ocean finfish and exchange ponds). We acknowledge that there are
probably important differences in potential vulnerability of individual
Methods species within these broad archetypes, and suggest that future research
Ethics declaration could usefully focus on understanding the diverse responses within
This study was granted exemption by the Ethics Committee of Xiamen groupings.
University. We certify that the study was performed in accordance with
the Declaration of Helsinki and later amendments. Informed consent Stressor identification
was obtained from all the participants (that is, all co-authors in this A scoping review was conducted to identify the key anthropogenic
study) before the study. stressors that have a direct impact on blue food production systems
(Supplementary Fig. 8a). Stressors impacting blue food production
Expert assessment protocol ‘quantity’ (volume produced) and ‘quality’ (food safety as it pertains
This study is part of Blue Food Assessment (www.bluefood.earth), an to human health) were addressed separately.
international joint initiative that brings together more than 100 scien- The most prevalent stressors impacting blue food production
tists from over 25 institutions spanning the globe. At the preparation quantity were first identified by screening published studies using
stage, a highly interdisciplinary group of over 20 experts was recruited the Google Scholar keyword searching tool. We used the search string
by the lead authors of this paper as co-authors, taking full account of (system name) AND (target name) AND (environment name) AND
the diversity of age, gender, cultural background and specialization. (extension 1 name) AND (extension 2 name), with detailed information
All co-authors are familiar with blue foods and each has expertise in provided in Supplementary Data 1. Each archetype corresponded to an
at least one area related to fisheries, aquaculture, seafood safety, cli- individual searching process. The first 500 items of each search process
mate change, environmental health or risk assessment. Following the (7,500 items in total for the 15 archetypes), sorted by relevance, were
standard 4-step principle (‘investigate–discuss–estimate–aggregate’) screened in titles, abstracts and keywords, yielding 508 useful articles
of expert elicitation46,47 (see flowchart in Supplementary Fig. 7), the within the search criteria. We acknowledge the limitations of keyword
overall aim was to establish an unstructured expert assessment proto- searching, which may not capture all relevant research, so a further
col to assist in study design and data selection and acquisition. Specific assessment of identified stressors was done by the author team to
objectives were to: (1) gather expert opinions on blue food system and check and verify the reasonableness of the stressors screened from
archetype classification, (2) select prevalent environmental stressors the literature and finally select the stressors on which all experts reach
and associated indicators and (3) determine vulnerability scores and consensus. In addition, stressors impacting blue food quality (that is,
uncertainty. It should be noted that in the latter two assessment pro- specific pollutants) were subject to an independent feasibility review
cesses, an integrated approach of scoping literature review and expert to assess the appropriate levels of evidence, drawing on the results of
assessment was used. Three expert assessments were conducted via a systematic review48.

After the above literature review was completed, we conducted a of each feature to a dimensionless range between 0 and 1. Different
second expert assessment with the co-author team. All experts were types of data were initially compiled at the national level as follows: (1)
first given the screening results from the literature and then asked two for geospatial data displayed in grids, data were aggregated to national
questions: “Which of the stressors are reasonable to be retained in this geographic boundaries (territorial land for inland systems and EEZ
research?” and “What are the indicators available for each stressor?”. On for marine systems) to calculate national averages using the ‘extract’
the basis of the literature and expert assessment, we identified a total of function in the ‘terra’ package in R; (2) for observation records with
17 well-documented key stressors impacting production quantity and spatial coordinates, data were converted into geospatial data using
food safety of blue foods. Considering geospatial data availability, we the ‘Kernel Density’ geoprocessing tool (with default parameters) in
further selected a globally available set of 28 indicators to represent and ArcGIS 10 and further calculated to obtain national averages; (3) for
quantify those stressors (see Supplementary Table 1 for data sources, observation records without spatial coordinates (for example, invasive
time span and resolution of all indicators). The indicators used to char- species and disease occurrence), data were counted and aggregated
acterize each stressor affecting the production quantity of marine and by country. In the complex blue food systems, there is no good way
inland blue food systems were slightly different. Sea surface tempera- to determine the weight of each indicator for each stressor, as it may
ture, pH and sea level rise were used as indicators of climate stressors vary from region to region. On the basis of our expert discussions, we
for marine systems, while altered precipitation and air temperature decided to assign the simplest equal weight to each indicator within
were used for inland systems. The incidence of floods, typhoons and each stressor to overcome the data limitation problem.
droughts was used as indicators of severe weather events. Hypoxia The exposure (E) of the blue food system (j) in a country (i) in
and eutrophication observations were used only for marine systems, response to a stressor (s) was calculated as a mean across indicators:
while nitrate and phosphorus usage were indicators of eutrophication n
for inland systems. Records of disease outbreaks and the number of 1
Ei,j,s = ∑ Xi,j,k × , (1)
invasive species found in each country, as well as the spatially recorded k=1
Ns
occurrence of lice, leeches and protozoans as indicators of parasites,
were obtained from public databases. Estimated mercury releases, where Xi,j,k is the normalized value of indicator k to which system
pesticide risks and antibiotic usage, as well as monitored concentra- j is exposed in the country i, and Ns is the number of indicators that
tion of polychlorinated biphenyls (PCBs), hexachlorocyclohexanes are used to create the composite measure for stressor s.
(HCHs) and dichloro-diphenyl-trichloroethane (DDT), were used to
reflect chemical stressors on food safety. The occurrence of HABs Vulnerability
was used to represent HAB toxins. The occurrence density of Listeria, The vulnerability of blue foods to environmental change was assessed
Salmonella, Vibrio and Yersinia was used to represent the major aquatic as a combination of sensitivity (that is, the degree to which a system is
foodborne bacteria. affected by environmental variability or change) and adaptive capac-
ity (that is, inherent ability to not only recover from but also adapt to
Countries included environmental change) using an integrated approach of scoping review
Countries and territories with records of blue food production in the (Supplementary Fig. 8b) and expert assessment. Review keyword plan-
most recent decade (between 2010 and 2019) were included in the ning and search string development were conducted in consultation
study (N = 222; 141 with both freshwater and marine production; 41 with with the co-author group. We used the search string (stressor name)
marine production only; and 40 with freshwater production only). Cur- AND (system name) AND (target name) AND (environment name) AND
rent blue food production data in the FAO database (https://www.fao. (extension 1 name) AND (extension 2 name), with detailed information
org/fishery/en/topic/166235) are not differentiated by archetype, so we provided in Supplementary Data 1. The included literature was limited
focused our spatial analysis on the four high-level production systems to publication (excluding books) after 2000 as a suitable cut-off both
(freshwater capture fisheries, freshwater aquaculture, marine capture for the inclusion of modern aquaculture systems and the inclusion of
fisheries and mariculture). In addition, production from brackish water increased rates of research related to many stressors. Each archetype–
was considered to be marine. stressor combination corresponded to a separate search process using
Google Scholar, comprising a total of 255 processes (15 archetypes
Exposure multiplied by 17 stressors). The first 200 results of each searching
We conducted a multifactor spatial analysis25,26 to assess the exposure process, sorted by relevance and screened in titles, abstracts and key-
of blue food production to environmental stresses by combining multi- words, yielded a total of 51,000 literature results. Full-text screening
ple indicators into a composite measure for each stressor. We assessed then determined inclusion in vulnerability assessments following the
exposure where the fishing activities happen within the national bound- criteria: (1) articles investigated an impact on production quantity or
aries of each nation and then aggregated that up to national-level quality (decrease or increase) explicitly linked to the relevant stressor;
statistics. This study excluded fishing in areas beyond national jurisdic- (2) articles demonstrated impact on a production system or production
tions because our focus on national-level results was intended to aid scenario (for example, specific farm, specific fishery) explicitly relevant
in policy and management decisions within jurisdictions. Historical to the archetype; (3) the impact on production was quantified at the
observation records (Supplementary Table 1) were used to reflect scale of the study, and in the case of food quality, an explicit link to a
the intensity of each indicator. Histograms for many of the indicators human health impact was made (although this might not have been
revealed skewness in their distributions (as detailed in Supplemen- quantified); (4) articles did not contain important methodological
tary Table 2), so these skewed indicators were first transformed (for flaws under the reviewers’ assessment; and (5) articles were published
example, logarithmic or square root transformation, see transform after the year 2000.
methods and effects in Supplementary Table 2) to support subsequent Due to data limitations, the vulnerability scoring was necessarily
analysis, which requires mean and variance as meaningful summaries a partially subjective process, so we conducted a third expert assess-
of the data. Outliers represent natural variations in this study, so they ment, with the co-authors divided into six groups and performing
were left for subsequent analysis. To make the variables measured at semi-quantitative and qualitative syntheses of the information related
different scales and units comparable for spatial analysis, we used the to archetypes, ranging from species to system level. Each expert group
min-max normalization method (formula: x’ = (x − min) / (max − min), was responsible for a different subset of archetype–stressor combina-
where x’ is the rescaled value, x is the original value, max is the maxi- tions. Each expert in the same group was asked to give an individual
mum value of x and min is the minimum value of x) to rescale the value score for each archetype–stressor assessment. Individual scores were

subsequently synthesized and shared with the entire co-author team (close to the expected value) and standard deviation (reflecting the
to allow for feedback until a consensus was reached. dispersion of the mean) of each indicator across all countries. Then
On the basis of the literature screening results, vulnerability levels we randomly simulated each indicator 1,000 times and recalculated
were qualitatively defined by co-authors as: (1) high: system perfor- the estimated total impact scores using the ‘rnorm’ function in R to
mance or safety severely impacted; (2) medium: system performance obtain the mean and standard deviation of the recalculated impact
or safety moderately impacted, or a moderate subsection of the sys- scores for each country.
tem was impacted; (3) low: low impact on performance at the system
level, or a minor subsection of the system was impacted; (4) none: no Cluster analysis
impact. The vulnerability of a high-level blue food production system To identify global patterns and find meaningful groupings of countries
to a specific stressor was determined as the mode value of archetype that shared similar impact or pressure profiles, a cluster analysis was
vulnerability levels in that system (Supplementary Table 3). The final performed on the basis of the outputs of the impact assessment. We
results were placed in a pre-coded Excel template. Key supporting used the conventional k-means algorithm53 due to its simplicity and
literature for each assessment is provided in Supplementary Data 2. inexpensiveness. The algorithm was employed to partition countries
Meanwhile, co-authors also discussed and quantified the uncer- into k distinct clusters such that each country belonged to the cluster
tainty for each vulnerability score for each archetype–stressor assess- with the ‘closest’ mean, where closeness was measured by the Euclid-
ment to communicate the level of our ‘confidence’ in the validity of the ean distance. Eight variables, including the estimated impact scores
result49,50. Our confidence scores were identified as appropriate by for both quantity and safety for all production systems, were used
experts and qualitatively defined as: (1) sufficient: a sufficient amount to calculate the Euclidean distances between clusters. We computed
of evidence existed for this assessment to be made, such as a relevant k-means clustering with the ‘kmeans’ function in R by varying k from
review. Impacts were communicated in the primary policy narratives 2 to 15 clusters. For each k, we calculated the total within-cluster sum
and the papers were largely up to date (<10 yr); (2) partial: a robust level of squares (WSS) and plotted the curve of the WSS (Supplementary
of evidence existed, but with gaps in knowledge or some difficulty in Fig. 10) to choose the best classification and determine the appropri-
aligning evidence directly with the archetype categorization. The ate groupings. All results were obtained with the algorithm using the
papers were relatively up to date (<20 yr); (3) poor: largely theoretical naïve method and default parameter settings. After determining the
evidence or poor empirical research that may only loosely align with appropriate number of clusters, we used the ‘pheatmap’ package in R to
the archetype categorization and few recent papers. (4) None: no create a heat map for visualization. Radar plots were used to show the
evidence was found. differences in the estimated impact on country clusters. This analysis
Alternatively, if finer-scale data become available, additional allows decision-makers to tailor mitigation efforts for each cluster
insights for adaptive strategies can be gained using a more in-depth to achieve greater effectiveness. Means and standard deviations of
analysis of vulnerability of archetypes to environmental stresses. As standardized impact scores for each country cluster are listed in Sup-
illustrated in our case study (Supplementary Fig. 9), we demonstrated plementary Table 4.
how to assess the expected consequence of anthropogenic pressures
on the production quantity of each blue food archetype in two related National response capacity
ways (exposure and vulnerability). Both were measured as a combina- We assessed the capacity (encompassing the three dimensions of
tion of the vulnerability of each archetype to a stressor, the produc- human and economic development and governance) of each blue food
tion amount of that archetype and the stressor intensity to which it nation to respond and implement effective adaptation strategies to
is exposed. cope with environmental stresses. Following ref. 54, we used the Human
Development Index (HDI) to represent human development, the Gross
Estimated impact Domestic Product (GDP) per capita to describe economic status and
The estimated impact or pressure of environmental change on blue the World Governance Index (WGI) to indicate governance effective-
food production was considered as a function of exposure and vulner- ness. These indicators are generalized socio-economic metrics and
ability 24. We calculated the estimated impact ( PIi,j ) as the are publicly available and continuously updated (see Supplementary
vulnerability-weighted sum of stressors for each stressor–system Table 5 and Fig. 11 for more details). All of the metrics were subjected to
combination, where the rescaled stressor intensity values were multi- min-max normalization that rescaled the data to values between 0 and
plied by the system’s categorical vulnerability score (Vs,j, 3.0 for high, 1, and the average of these rescaled metrics was then used to denote
2.0 for medium, 1.0 for low, 0 for none; see Supplementary Table 2) to the response capacity score of each nation54.
the stressor.
s
Bonferroni outlier test
PIi,j = ∑ Ei,j,s × Vs,j , (2) The relationship between estimated impacts and national response
1 capacity at the country level was explored on the basis of a generalized
linear model. We also aimed to test all observations simultaneously
where Vs,j is the categorical vulnerability score of system j to stressor to identify country outliers, so we used the Bonferroni outlier test to
s, and Ei,j,s is the exposure of system j in country i to stressor s. protect from type I error55. We used the ‘outlierTest’ function in the ‘car’
To facilitate mapping, the estimated impact scores were first package in R to perform the test. Countries with Bonferroni P < 0.05 were
rescaled using min-max normalization, such that each country had a considered mean-shift outliers that had extreme observations furthest
value from 0 to 1 for each system, and then categorized into quintiles from the mean value of the neighbourhood (Supplementary Table 6).
and mapped at the national level in R.
Reporting summary
Uncertainty analysis Further information on research design is available in the Nature Port-
As our results were highly dependent on data availability and the quality folio Reporting Summary linked to this article.
of reported data, we further included an uncertainty analysis to show
how the expected outcomes of environmental change vary when jit- Data availability
tering the stress indicators. National exposure to each stress indicator Source data were all retrieved from publicly accessible databases as
is the sum of uncertain terms51,52. On the basis of the indicator infor- presented in Supplementary Tables 1 and 5. Additional data are avail-
mation listed in Supplementary Table 2, we first calculated the mean able in the Supplementary Data file.

Code availability 24. Butt, N. et al. A trait-based framework for assessing the
The analysis code used to produce the results of this study is available at vulnerability of marine species to human impacts. Ecosphere 13,
https://github.com/bluefoodvulnerability/bluefoodvulnerability.git. e3919 (2022).
25. Biotto, G. et al. GIS, multi-criteria and multi-factor spatial analysis
References for the probability assessment of the existence of illegal landfills.
1. Costello, C. et al. The future of food from the sea. Nature 588, Int. J. Geogr. Inf. Sci. 23, 1233–1244 (2009).
95–100 (2020). 26. Shan, N., Ruan, X. & Ao, J. 2010 18th International Conference on
2. The State of World Fisheries and Aquaculture 2020. Sustainability Geoinformatics (IEEE, 2010).
in Action (FAO, 2020); http://www.fao.org/documents/card/en/c/ 27. Bender, E. A., Case, T. J. & Gilpin, M. E. Perturbation experiments
ca9229en.pdf in community ecology: theory and practice. Ecology 65,
3. Golden, C. D. et al. Aquatic foods to nourish nations. Nature 598, 1–13 (1984).
315–320 (2021). 28. Brown, A. R. et al. Assessing risks and mitigating impacts of
4. Naylor, R. L. et al. Blue food demand across geographic and harmful algal blooms on mariculture and marine fisheries.
temporal scales. Nat. Commun. 12, 5413 (2021). Rev. Aquac. 12, 1663–1688 (2020).
5. Short, R. E. et al. Harnessing the diversity of small-scale 29. IPCC Climate Change 2007: Impacts, Adaptation and
actors is key to the future of aquatic food systems. Nat. Food 2, Vulnerability (eds Parry, M. L. et al.) 779–810 (Cambridge
733–741 (2021). Univ. Press, 2007).
6. Teh, L. C. L. & Sumaila, U. R. Contribution of marine fisheries to 30. Canham, R., González‐Prieto, A. M. & Elliott, J. E. Mercury
worldwide employment: global marine fisheries employment. exposure and toxicological consequences in fish and fish‐eating
Fish Fish. 14, 77–88 (2013). wildlife from anthropogenic activity in Latin America. Integr.
7. Gephart, J. A. et al. Environmental performance of blue foods. Environ. Assess. Manage. 17, 13–26 (2021).
Nature 597, 360–365 (2021). 31. Hadjimichael, C., Georgiou, K., Samoutis, G. & Demetriades, E.
8. Gordon, L. J. et al. Rewiring food systems to enhance human Sales of systemic anti-infective agents in Cyprus in comparison
health and biosphere stewardship. Environ. Res. Lett. 12, with four other European countries. Pharm. World Sci. 28,
100201 (2017). 135–139 (2006).
9. Poore, J. & Nemecek, T. Reducing food’s environmental impacts 32. Sanches, F. H. C. et al. Aggressiveness overcomes body-size
through producers and consumers. Science 360, 987–992 (2018). effects in fights staged between invasive and native fish species
10. Tilman, D. & Clark, M. Global diets link environmental with overlapping niches. PLoS ONE 7, e29746 (2012).
sustainability and human health. Nature 515, 518–522 (2014). 33. Chessman, B. C. Identifying species at risk from climate change:
11. Troell, M. et al. Does aquaculture add resilience to the global food traits predict the drought vulnerability of freshwater fishes.
system? Proc. Natl Acad. Sci. USA 111, 13257–13263 (2014). Biol. Conserv. 160, 40–49 (2013).
12. Willett, W. et al. Food in the Anthropocene: the EAT–Lancet 34. Bell, J. D. et al. Mixed responses of tropical Pacific fisheries
Commission on healthy diets from sustainable food systems. and aquaculture to climate change. Nat. Clim. Change 3,
Lancet 393, 447–492 (2019). 591–599 (2013).
13. Bennett, A. et al. Recognize fish as food in policy discourse and 35. Hovel, R. A., Thorson, J. T., Carter, J. L. & Quinn, T. P. Within‐lake
development funding. Ambio 50, 981–989 (2021). habitat heterogeneity mediates community response to warming
14. Halpern, B. S. et al. Recent pace of change in human impact on trends. Ecology 98, 2333–2342 (2017).
the world’s ocean. Sci. Rep. 9, 11609 (2019). 36. Castilhos, Z. C. et al. Mercury contamination in fish from gold
15. Cottrell, R. S. et al. Considering land–sea interactions and mining areas in Indonesia and human health risk assessment. Sci.
trade-offs for food and biodiversity. Glob. Change Biol. 24, Total Environ. 368, 320–325 (2006).
580–596 (2018). 37. IPCC Special Report on Managing the Risks of Extreme Events and
16. Blanchard, J. L. et al. Linked sustainability challenges and Disasters to Advance Climate Change Adaptation (Cambridge
trade-offs among fisheries, aquaculture and agriculture. Univ. Press, 2012).
Nat. Ecol. Evol. 1, 1240–1249 (2017). 38. Jurgilevich, A., Räsänen, A., Groundstroem, F. & Juhola, S. A
17. The State of World Fisheries and Aquaculture 2018. Meeting the systematic review of dynamics in climate risk and vulnerability
Sustainable Development Goals (FAO, 2018); https://www.fao. assessments. Environ. Res. Lett. 12, 013002 (2017).
org/3/i9540en/i9540en.pdf 39. Tyberghein, L. et al. Bio‐ORACLE: a global environmental dataset
18. Froehlich, H. E., Gentry, R. R. & Halpern, B. S. Global change in for marine species distribution modelling. Glob. Ecol. Biogeogr.
marine aquaculture production potential under climate change. 21, 272–281 (2012).
Nat. Ecol. Evol. 2, 1745–1750 (2018). 40. Assis, J. et al. Bio‐ORACLE v2.0: extending marine data
19. Tigchelaar, M. et al. Compound climate risks threaten aquatic layers for bioclimatic modelling. Glob. Ecol. Biogeogr. 27,
food system benefits. Nat. Food 2, 673–682 (2021). 277–284 (2018).
20. Cao, L. et al. China’s aquaculture and the world’s wild fisheries. 41. Handisyde, N., Telfer, T. C. & Ross, L. G. Vulnerability of
Science 347, 133–135 (2015). aquaculture-related livelihoods to changing climate at the global
21. Clavelle, T., Lester, S. E., Gentry, R. & Froehlich, H. E. Interactions scale. Fish Fish. 18, 466–488 (2017).
and management for the future of marine aquaculture and 42. Joffre, O. M., Poortvliet, P. M. & Klerkx, L. Are shrimp farmers
capture fisheries. Fish Fish. 20, 368–388 (2019). actual gamblers? An analysis of risk perception and risk
22. Naylor, R. L. et al. Effect of aquaculture on world fish supplies. management behaviors among shrimp farmers in the Mekong
Nature 405, 1017–1024 (2000). Delta. Aquaculture 495, 528–537 (2018).
23. HLPE Agroecological and Other Innovative Approaches for 43. McConney, P., Cox, S. A. & Parsram, K. Building food security and
Sustainable Agriculture and Food Systems that Enhance Food resilience into fisheries governance in the Eastern Caribbean.
Security and Nutrition (Committee on World Food Security, Reg. Environ. Change 15, 1355–1365 (2015).
2019); https://www.fao.org/fileadmin/user_upload/hlpe/hlpe_ 44. Ding, Q., Chen, X., Hilborn, R. & Chen, Y. Vulnerability to impacts
documents/HLPE_S_and_R/HLPE_2019_Agroecological- of climate change on marine fisheries and food security.
and-Other-Innovative-Approaches_S-R_EN.pdf Mar. Policy 83, 55–61 (2017).

45. Hobday, A. J. et al. Planning adaptation to climate change in Author contributions

fast-warming marine regions with seafood-dependent coastal L.C., B.S.H. and M. Troell designed and conceptualized the research
communities. Rev. Fish. Biol. Fish. 26, 249–264 (2016). idea. Scoping review was performed by L.C., B.S.H., M. Troell, R.S.,
46. Hanea, A. M. et al. I nvestigate D iscuss E stimate A ggregate for C. Zeng, Z.J., Y.L., C. Zou, C.L., S.L., X.L., W.W.L.C., R.S.C., F.D., S.G.,
structured expert judgement. Int. J. Forecast. 33, 267–279 (2017). J.A.G., D.G.-S., J.I.K., F.M., R.L.N., H.J.P., E.R.S., U.R.S. and M. Tigchelaar.
47. Hemming, V., Burgman, M. A., Hanea, A. M., McBride, M. F. & Data acquisition was done by L.C., M. Troell, R.S., C. Zeng, Z.J., Y.L.,
Wintle, B. C. A practical guide to structured expert elicitation C. Zou, C.L., S.L. and X.L. Data compilation, spatial analysis and data
using the IDEA protocol. Methods Ecol. Evol. 9, 169–180 (2018). interpretation were conducted by L.C., C. Zeng, Z.J. and Y.L. The
48. Short, R. E. et al. Review of the evidence for oceans and human original paper was drafted by L.C., B.S.H., M. Troell, R.S. and Z.J., and
health relationships in Europe: a systematic map. Environ. Int. 146, edited and revised by L.C., B.S.H., M. Troell, R.S., C. Zeng, Z.J., Y.L.,
106275 (2021). C. Zou, C.L., S.L., X.L., W.W.L.C., R.S.C., F.D., S.G., J.A.G., D.G.-S., J.I.K.,
49. Gardali, T., Seavy, N. E., DiGaudio, R. T. & Comrack, L. A. A climate F.M., R.L.N., H.J.P., E.R.S., U.R.S. and M. Tigchelaar.
change vulnerability assessment of California’s at-risk birds.
PLoS ONE 7, e29507 (2012). Competing interests
50. Case, M. J. & Lawler, J. J. Relative vulnerability to climate The authors declare no competing interests.
change of trees in western North America. Clim. Change 136,
367–379 (2016). Additional information
51. Syri, S., Suutari, R. & Posch, M. From emissions in Europe to Supplementary information The online version
critical load exceedances in Finland—uncertainty analysis contains supplementary material available at
of acidification integrated assessment. Environ. Sci. Policy 3, https://doi.org/10.1038/s41893-023-01156-y.
263–276 (2000).
52. Pardo, D., Jenouvrier, S., Weimerskirch, H. & Barbraud, C. Effect of Correspondence and requests for materials should be addressed
extreme sea surface temperature events on the demography of to Ling Cao.
an age-structured albatross population. Phil. Trans. R. Soc. B 372,
20160143 (2017). Peer review information Nature Sustainability thanks Anna Farmery,
53. Hartigan, J. A. & Wong, M. A. Algorithm AS 136: a k-means Kristy Kroeker, Pauline Scheelbeek and the other, anonymous,
clustering algorithm. J. R. Stat. Soc. C 28, 100–108 (1979). reviewer(s) for their contribution to the peer review of this work.
54. Varis, O., Taka, M. & Kummu, M. The planet’s stressed river basins:
too much pressure or too little adaptive capacity? Earth’s Future 7, Reprints and permissions information is available at
1118–1135 (2019). www.nature.com/reprints.
55. Wehde, W. et al. Quantitative Research Methods for Political
Science, Public Policy and Public Administration for Publisher’s note Springer Nature remains neutral with regard to
Undergraduates: 1st Edition With Applications in Excel jurisdictional claims in published maps and institutional affiliations.
(East Tennessee State University, 2020).
Open Access This article is licensed under a Creative Commons
Acknowledgements Attribution 4.0 International License, which permits use, sharing,
This paper is part of Blue Food Assessment (https://www.bluefood. adaptation, distribution and reproduction in any medium or format,
earth/), a comprehensive examination of the role of aquatic foods as long as you give appropriate credit to the original author(s) and the
in building healthy, sustainable and equitable food systems. source, provide a link to the Creative Commons license, and indicate
The assessment has benefitted from the intellectual input of the if changes were made. The images or other third party material in this
wider group of scientists leading other components of the BFA article are included in the article’s Creative Commons license, unless
work. We thank the Ministry of Science and Technology of China indicated otherwise in a credit line to the material. If material is not
(2022YFC3102404 to L.C.), the Shanghai Pilot Program for Basic included in the article’s Creative Commons license and your intended
Research-Shanghai Jiao Tong University (21TQ1400220 to L.C. and C. use is not permitted by statutory regulation or exceeds the permitted
Zeng), the National Science Foundation of China (42142018 to L.C.) use, you will need to obtain permission directly from the copyright
and the David and Lucile Packard Foundation (to L.C.) for financial holder. To view a copy of this license, visit http://creativecommons.
support. Any opinions, findings, conclusions or recommendations org/licenses/by/4.0/.
expressed in this material are those of the authors and do not
necessarily reflect the views of the funders. © The Author(s) 2023
State Key Laboratory of Marine Environmental Science, College of Ocean and Earth Sciences, Xiamen University, Xiamen, P.R. China. 2National Center for
1
Ecological Analysis and Synthesis, University of California, Santa Barbara, CA, USA. 3Bren School of Environmental Science and Management, University
of California, Santa Barbara, CA, USA. 4Stockholm Resilience Centre, Stockholm University, Stockholm, Sweden. 5Beijer Institute, Royal Swedish Academy
of Sciences, Stockholm, Sweden. 6School of Oceanography, Shanghai Jiao Tong University, Shanghai, P.R. China. 7Institute for the Oceans and Fisheries,
University of British Columbia, Vancouver, British Columbia, Canada. 8Institute of Marine and Antarctic Studies, University of Tasmania, Hobart, Tasmania,
Australia. 9EAT, Oslo, Norway. 10Instituto Milenio en Socio-Ecologia Costera and Center of Applied Ecology and Sustainability, Facultad de Ciencias
Biologicas, Pontificia Universidad Catolica de Chile, Santiago, Chile. 11Department of Environmental Science, American University, Washington, DC,
USA. 12School of Earth, Energy and Environmental Sciences, Stanford University, Stanford, CA, USA. 13Stanford Center for Ocean Solutions and Oceans
Department, Hopkins Marine Station, Stanford University, Pacific Grove, CA, USA. 14Department of Earth System Science and Center on Food Security
and the Environment, Stanford University, Stanford, CA, USA. 15Center for Ocean Solutions, Stanford University, Stanford, CA, USA. 16School of
Public Policy and Global Affairs, University of British Columbia, Vancouver, British Columbia, Canada. 17These authors contributed equally: Ling Cao,
Benjamin S. Halpern, Max Troell, Rebecca Short, Cong Zeng, Ziyu Jiang, Yue Liu. e-mail: caoling@xmu.edu.cn

nature research | reporting summary
Corresponding author(s): Ling Cao
Last updated by author(s): 5/16/2023
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 (n) 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
Only common tests should be described solely by name; describe more complex techniques in the Methods section.
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. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.
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 d, Pearson's r), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.
Software and code

Policy information about availability of computer code
Data collection No software was used for data collection.
Data analysis All data analysis steps are described in Methods. Briefly, when compiling data of environmental stressors at the national level, the extract
function in terra package (version 1.6-7) in R (version 4.1.1) was used for geospatial data displayed in grids, and the Kernel Density tool in
ArcGIS (version 10.2.2) was used for observation records with spatial coordinates. The rnorm function in R (version 4.1.1) was used to obtain
the mean and standard deviation of the recalculated impact scores in uncertainty analysis. The kmeans function in R (version 4.1.1) was used
to compute k-means clustering. The pheatmap package (version 1.0.12) in R (version 4.1.1) was used for visualization of cluster analysis. The
outlierTest function in car package (version 3.1-1) in R (version 4.1.1) was used for Bonferroni outlier test. The essential custom code
developed for this study is available in Github at: https://github.com/bluefoodvulnerability/bluefoodvulnerability.git.
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
Source data were all retrieved from publicly accessible databases as presented in Table S1 and Table S5. Additional data are available in the Supplementary Data file.
Publicly accessible databases include the following:
1
[1] FAO. FishStatJ. URL https://www.fao.org/fishery/en/topic/166235.

[2] University of East Anglia. CRU datasets. URL https://crudata.uea.ac.uk.
[3] NOAA. NCEI metadata. URL https://www.ncei.noaa.gov/access/metadata/landing-page/bin/iso?id=gov.noaa.nodc:0206289.
[4] NOAA. PSL data. URL https://psl.noaa.gov/data/gridded/data.gpcp.html.
[5] Aviso+. Aviso data products. URL https://www.aviso.altimetry.fr/en/data/products/ocean-indicators-products/mean-sea-level/data-acces.html#c12195.
[6] NCAR. IBtrACS: Tropical cyclone best track data. URL https://climatedataguide.ucar.edu/climate-data/ibtracs-tropical-cyclone-best-track-data.
[7] EM-DAT. EM-DAT data. URL https://public.emdat.be/data.
[8] GBIF. GBIF occurrence download. URL https://www.gbif.org.
[9] World Organization of Animal Health. Quantitative data dashbord. URL https://wahis.oie.int/#/dashboards/qd-dashboard.
[10] IUCN. Global Invasive Species Database. URL http://www.iucngisd.org/gisd.
[11] WRI. Eutrophication & Hypoxia Map Data Set. URL https://www.wri.org/data/eutrophication-hypoxia-map-data-set.
[12] Our World in Data. Data charts. URL https://ourworldindata.org/charts.
[13] International Pellet Watch. Global pollution map. URL http://www.pelletwatch.org.
[14] IOC, UNESCO. Harmful Algae Event Database. URL http://haedat.iode.org/browseGrids.php.
[15] UNDP. Human Development Reports. URL http://hdr.undp.org/en/data.
[16] World Bank. Worldwide Governance Indicators. URL https://info.worldbank.org/governance/wgi.
[17] World Bank. DataBank. URL https://databank.worldbank.org.
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
Ecological, evolutionary & environmental sciences study design

All studies must disclose on these points even when the disclosure is negative.
Study description Vulnerability analysis of blue foods under the stress of anthropogenic environmental change
Research sample This study involves existing datasets. See above for the publicly available data sources.
Sampling strategy N/A
Data collection Publicly available data were used in the study and web links were provided as above.
Timing and spatial scale Global scale.
Data exclusions N/A
Reproducibility All results are reproducible via the code provided.
Randomization N/A
Blinding N/A
Did the study involve field work? Yes No
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
April 2020
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
2
Human research participants

Policy information about studies involving human research participants
Population characteristics This study enlists the expertise of its co-authors, all of whom are human participants, to evaluate the environmental
pressures on blue food production. The participant pool is diverse and not defined by characteristics such as race, ethnicity,
national or social origin, sex, or gender identity.
Recruitment At the preparation stage, a highly interdisciplinary group of over 20 experts was recruited by the lead authors of this paper as
coauthors. All coauthors are familiar with blue foods and each has expertise in at least one area related to fisheries,
aquaculture, seafood safety, climate change, environmental health, or risk assessment.
Ethics oversight This study was granted exemption by the Ethics Committee of Xiamen University. We certify that the study was performed in
accordance with the Declaration of Helsinki and later amendments. Informed consent was obtained from all the participants
(i.e., all co-authors in this study) prior to the study.
Note that full information on the approval of the study protocol must also be provided in the manuscript.
April 2020

Vulnerability of Blue Foods To Human-Induced Environmental Change

Uploaded by

Copyright:

Available Formats

Vulnerability of Blue Foods To Human-Induced Environmental Change

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

Vulnerability of Blue Foods To Human-Induced Environmental Change

Uploaded by

Copyright:

Available Formats

nature sustainability

Vulnerability of blue foods to

Nature Sustainability | Volume 6 | October 2023 | 1186–1198 1186

Nature Sustainability | Volume 6 | October 2023 | 1186–1198 1187

UK Trinidad and Tob.

Cambodia Trinidad and Tob.

Bosnia and Herz.

0 0.25 0.50 0.75 1.00 0 0.25 0.50 0.75 1.00

Nature Sustainability | Volume 6 | October 2023 | 1186–1198 1188

FW capture FW culture Marine capture Mariculture

Sea level rise

Severe weather events

Nature Sustainability | Volume 6 | October 2023 | 1186–1198 1189

Nature Sustainability | Volume 6 | October 2023 | 1186–1198 1190

Standardized impact score

0.0–0.2 0.2–0.4 0.4–0.6 0.6–0.8 0.8–1.0

Marine capture Mariculture

Marine capture Mariculture

Nature Sustainability | Volume 6 | October 2023 | 1186–1198 1191

Standardized impact score

Mariculture safety FW culture quantity

Marine capture safety 0 FW capture safety

Mariculture quantity FW culture safety

Marine capture quantity

Medium low impact Low impact

Nature Sustainability | Volume 6 | October 2023 | 1186–1198 1192

0 0.25 0.50 0.75 1.00 0 0.25 0.50 0.75 1.00

1.00 Denmark 1.00

0.75 Japan 0.75

0 0.25 0.50 0.75 1.00 0 0.25 0.50 0.75 1.00

1.00 Albania 1.00

0 0.25 0.50 0.75 1.00 0 0.25 0.50 0.75 1.00

Bosnia and Herz.

0 0.25 0.50 0.75 1.00 0 0.25 0.50 0.75 1.00

National response capacity

Nature Sustainability | Volume 6 | October 2023 | 1186–1198 1193

Nature Sustainability | Volume 6 | October 2023 | 1186–1198 1194

Nature Sustainability | Volume 6 | October 2023 | 1186–1198 1195

Nature Sustainability | Volume 6 | October 2023 | 1186–1198 1196

Nature Sustainability | Volume 6 | October 2023 | 1186–1198 1197

45. Hobday, A. J. et al. Planning adaptation to climate change in Author contributions

Nature Sustainability | Volume 6 | October 2023 | 1186–1198 1198

A description of all covariates tested

Software and code

Policy information about availability of data

nature research | reporting summary

Ecological, evolutionary & environmental sciences study design

Sampling strategy N/A

Timing and spatial scale Global scale.

Data exclusions N/A

Reproducibility All results are reproducible via the code provided.

Did the study involve field work? Yes No

Reporting for specific materials, systems and methods

Materials & experimental systems Methods

Eukaryotic cell lines Flow cytometry

nature research | reporting summary

You might also like