The food system is a major driver of environmental change, emitting a quarter of all greenhouse gas (GHG) emissions, occupying half of all ice-free land, and responsible for three quarters of global consumptive water use and eutrophication3,6. Yet, it still fails to meet global nutrition needs7, with 820 million people lacking sufficient food8 and with one in three people globally overweight or obese9. As a critical source of nutrition8,10 generating relatively low average environmental pressures1,2,11,12, blue foods present an opportunity to improve nutrition with lower environmental burdens, in line with the Sustainable Development Goals to improve nutrition (Goal 2), ensure sustainable consumption and production (Goal 12), and sustainably use marine resources (Goal 14).

Blue foods, however, are underrepresented in food system environmental assessments13 and the stressors considered are limited4 such that we have some understanding of GHG emissions14,15, but less of others such as land or freshwater use16. Where blue foods are included, they are typically represented by only one or a few broad categories (see, for example, refs. 3,17,18), masking the vast diversity within blue food production. Finally, estimates combining results of published life cycle assessments undertaken for different purposes, and consequently using incompatible methodologies19,20, cannot be compared reliably. It is therefore critical to examine the environmental performance across the diversity of blue foods in a robust, methodologically consistent manner to serve as a benchmark within the rapidly evolving sector as blue food demand increases21, production shifts toward aquaculture and production technologies advance.

Here, we provide standardized estimates of GHG emissions, consumptive freshwater use (water use), terrestrial land occupation (land use), and nitrogen (N) and phosphorus (P) emissions for blue foods, reported per tonne of edible weight. We identify a set of key life cycle inventory data (that is, material and energy input, and farm-level performance data) from published studies and datasets to which a harmonized methodology is applied. We draw on studies that collectively report data from over 1,690 farms and 1,000 unique fishery records around the world. The 23 species groups represented in our results cover over 70% of global blue food production. We then discuss environmental impacts not covered by the standard stressors, most notably biodiversity loss. Finally, we leverage our model to identify and quantify improvement opportunities and discuss public and private policy options to realize these improvements. In doing so, these results help to identify current and future opportunities for blue foods within sustainable diets.

Blue food environmental stressors

Reducing food system GHG emissions is central to meeting global emission targets8. Fed aquaculture emissions result primarily from feeds22, while fuel use drives capture fisheries emissions11. Across assessed blue foods, farmed seaweeds and bivalves generate the lowest emissions, followed by small pelagic capture fisheries, while flatfish and crustacean fisheries produce the highest (Fig. 1). For fed aquaculture, feed production is responsible for more than 70% of emissions for most groups (Supplementary Fig. 6). Farmed bivalves and shrimp produce lower average emissions than their capture counterparts (bivalves, 1,414 versus 11,400 kgCO2e t−1 (kilograms of CO2 equivalent per tonne); shrimps, 9,428 versus 11,956 kgCO2e t−1), while salmon/trout are similar whether farmed or fished (5,101–5,410 versus 6,881 kgCO2e t−1).

Fig. 1: Stressor posterior distributions.
figure 1

a, Aquaculture GHG emissions (kgCO2e t−1). b, Aquaculture N (kgNe t−1). c, Aquaculture P (kgPe t−1). d, Capture GHG emissions (kgCO2e t−1) e, Aquaculture Water use (m3 t−1). f, Aquaculture land use (m2a t−1). Values represent tonnes of edible weight and use mass allocation. Dot indicates the median, coloured regions show credible intervals (that is, range of values that have a 95% (light), 80% and 50% (dark) probability of containing the true parameter value). Beige bands represent estimated chicken minimum to maximum range. See Supplementary Fig 10 for estimates expressed in terms of live weight Source data.

Land use, especially conversion of natural areas, results in a range of context-dependent biodiversity impacts and GHG emissions23 and creates potential trade-offs with alternate uses, including production of other foods. On-farm land use is low (<1,000 m2 annual terrestrial land occupation per tonne, m2a t−1; <10%) for most systems and highest (3,737–8,689 m2a t−1) for extensive ponds (for example, milkfish, shrimp and silver and bighead carp). Generally, most land use is associated with feed production for fed systems and explains the overall rankings (Fig. 1), though milkfish uses the highest amount of on- and off-farm land.

Freshwater increasingly constrains agriculture production but capture fisheries and unfed mariculture require little to no freshwater24. Although blue foods are produced in water, consumptive freshwater use is largely limited to feed production and on-farm evaporative losses for freshwater production16, with feeds accounting for essentially all water use for marine species, but on-farm evaporative losses accounting for over 60% of water use for freshwater species. (Supplementary Fig. 6). High evaporative losses cause silver and bighead carps to have the highest total water use, 2.6 times the water use of other carps and 4.4 times the water use of catfish, while milkfish and miscellaneous marine and diadromous fishes have the highest feed-associated water use. Among fed aquaculture, trout and salmon have the lowest water use, in part attributable to lower crop utilization, highlighting a trade-off with fishmeal and fish oil.

Nitrogen and phosphorus emissions are responsible for marine and freshwater eutrophication and are highly correlated due to natural biomass N:P ratios (Supplementary Table 4). For fed systems, the majority of N (>87%) and P (>94%) emissions occur on-farm. The highest total N and P emissions result from miscellaneous farmed marine and diadromous fishes, milkfish and fed carp. Non-fed groups such as seaweeds and bivalves, as well as unfed and unfertilized finfish systems (for example, some silver and bighead carp), represent extractive systems that remove more N and P than is emitted during production, resulting in negative emissions (Fig. 1).

Across all blue foods, farmed seaweeds and bivalves generate the lowest stressors. However, these groups also highlight several assumptions and nuances. First, bivalve estimates change by nearly five-fold when expressed in terms of edible portion (Fig. 1) compared to live weight (Supplementary Fig. 10) due to the shell weight. Second, some processes falling outside our system boundaries represent a potentially large fraction of life cycle emissions for these groups, even if still small in absolute value in some cases. For seaweeds, a large proportion of GHG emissions can occur at the drying stage25 while for bivalves, CO2 emissions during shell formation26 and high emissions associated with live product from transport27 can be important. Third, impacts on biogeochemical cycling and habitats are highly context dependent. For example, the systems represented here extract nitrogen and phosphorus, which could be problematic in nutrient-poor environments. Additionally, ozone effects from volatile short-lived substances depend on the location and varies widely across species28,29. Fourth, sustainable diet recommendations based on these or similar results must account for differences in nutrition content and bioavailability, a particularly important consideration for seaweeds30. Finally, these systems are underrepresented in the literature, particularly for edible seaweeds (Supplementary Fig. 3). As recommendations point towards the potential of these groups, it is important to increase data on these systems, deepen understanding of the above nuances, and be mindful of the total impacts associated with large-scale production on coastal habitats.

Capture fisheries, with negligible land, water, N and P values, also compare favourably, though groups fall at both the bottom and top of GHG rankings. Among farmed finfish and crustaceans, silver and bighead carps result in the lowest GHG, N and P emissions, while salmon and trout use the least land and water. To compare with terrestrial foods, we estimated stressors for industrial chicken produced in the USA and Europe and find it falls in the middle of farmed blue foods, with similar stressors as tilapia (Fig. 1, Supplementary Fig. 14). Because chicken typically has lower stressors than other livestock3, it follows that many blue food groups compare favourably to other animal-sourced foods. Notably, groups generating among the lowest stressors (for example, bivalves and small pelagic fishes) also provide the greatest nutritional quality across all forms of aquatic foods2,10.

Our results represent the most comprehensive and standardized blue food stressor estimates to date. Overall, data availability is correlated with global aquaculture production across these taxa groups, but there are still notable taxonomic and geographic gaps (Supplementary Figs. 3, 4). Critically, there are substantial data gaps for silver and bighead carp and seaweeds given their level of production (Supplementary Fig. 3). Furthermore, our capture fishery data primarily represents commercial marine fisheries31. However, subsistence marine and inland catches often utilize non‐motorized or no vessels, which probably generate few emissions, but there is insufficient data on fuel use across the diversity of small-scale fishing methods to reliably estimate emissions. These systems should be prioritized for additional research. Our estimates represent a snapshot of the knowledge of current production, but future work on emerging production technologies, feed innovations and growing sub-sectors is important for tracking changes against these benchmarks.

From stressors to ecosystem impacts

Emission and resource-use stressors are valuable for comparing environmental performance across foods but cannot fully capture final ecosystem and biodiversity consequences (that is, impacts). Estimating impacts stemming from blue food production requires considering additional stressors and accounting for local context.

While GHG, N, P, land and water are important stressors commonly used to compare foods, other less studied stressors can be critical drivers of ecosystem impacts (Fig. 2). Both aquaculture and fisheries may impose other stressors through toxic substance applications (for example, antifouling and pesticides in agriculture) and physical disturbance (for example, bottom trawling and on-bottom culture). Additional stressors include genetic pollution, invasive species introductions32, application of antibiotics33, and disease spread34. While capture fisheries have negligible N, P, water and land stressors, other stressors can markedly alter ecosystems. Fisheries often shift size structure and abundance of targeted species (see, for example, refs. 35,36), alter the structure of food webs (see, for example, ref. 37) and impact non-targeted fauna through bycatch38.

Fig. 2: Major stressors stemming from aquaculture and capture fisheries.
figure 2

Icons with magenta border are quantified in this study while the others are discussed qualitatively.

Local context, such as ecosystem function, carrying/assimilating capacity, and species composition influence how stressors translate into environmental impacts39,40. Notably, land use impacts on biodiversity depend on the land use history and ecological context41. While all land used for food cultivation represents habitats converted at one point, avoiding additional agricultural expansion is important for preventing further habitat loss42. This is also true for on-farm land use by aquaculture, where conversion of ecologically valuable ecosystems, such as mangrove forests23 that serve as critical carbon sinks43 and nursery habitats, can generate severe impacts. Local species composition and management contexts are also important, including risks associated with marine mammal bycatch (Box 1). Individual stressors may also have nonlinear relationships with impacts or act interactively44,45, such as climate change impacts compounding land use patterns that limit climate refuges or migration options46, or resulting in more frequent disease outbreaks, that increase antibiotic use and risk of antibiotic resistance.

Fig. 3: GHG emissions compared to marine mammal risk.
figure 3

Data represent fisheries in Europe (NE Atl) and Central America (C Am SSF) by gear type. Dot indicates the median estimate of the mean kgCO2e t−1 and intervals show 95% (light), 80% and 50% (dark) credible intervals. Risk index is the sum of the number of marine mammals at risk times 3, 2 or 1 for high, medium or low risk, respectively Source data.

Capturing the full suite of environmental impacts will require more systematic data collection and methodological advancements. This is crucial for informing policy decisions and realizing the potential contributions of blue foods to sustainable diets while avoiding undesirable trade-offs. Combining local ecological risk and stressor estimates can reveal these important trade-offs, as well as potential synergies (Box 1). While there are no impact-free foods, highlighting synergies simplifies sustainability messaging and helps identify priority interventions.

Levers for reducing environmental impacts

Variance in stressors indicates diversity across fishing/farming systems (Supplementary Figs. 79) as well as potential ‘performance gaps’. High variability in milkfish and miscellaneous marine and diadromous fish stressors points to large potential performance gains per unit. This is promising given the interest in marine finfish expansion47. Meanwhile, smaller performance gains per unit for high production groups such as carps are likely to generate larger total gains. While some variability within a taxa group is due to differences in on-farm practices, production technology is an important factor across stressors48 as variability in stressors for a given species reared in different farming systems can be considerable (see, for example, ref. 49).

We find feed conversion ratios (FCRs) represent the strongest lever, wherein a 10% reduction results in a 1–24% decrease in all stressors (Fig. 4a). To evaluate potential shifts under current technology, we estimate the effect of moving each species to the 20th percentile FCR and find the largest reductions for silver and bighead carps (Fig. 4b). However, lower FCRs generally come at the cost of larger pond area33, suggesting a potential trade-off with land and water use.

Fig. 4: Aquaculture stressor intervention opportunities.
figure 4

a, Change (%) in each stressor associate with a 10% reduction in the parameter value (black cell indicates stressor change >20%). b, Change (%) in each stressor under four scenarios (defined in Supplementary Table 8) relative to the current estimate. Arrows indicate changes greater than 50%. Additional aquaculture scenario results displayed in Supplementary Fig. 15 and capture scenario results in Supplementary Fig. 16 Source data.

Holding all else constant, a 10% fish production yield improvement (t ha−1) reduces land and water use for freshwater pond systems by 1–10% (Fig. 4a). Increasing yields to the 80th percentile reduces land and water use by up to 50% (Fig. 4b). Intensifying production, however, can require more energy for aeration and water pumping as well as increased disease risks with higher animal densities.

Feed composition represents another potential lever. Overall, shifting relative proportions of crop- and fish-derived inputs to feeds results in negligible changes in stressors (Fig. 4a). Comparing changes in feed sourcing, we found switching to deforestation-free soy and crops reduced GHG emissions by 5–50% (Fig. 4b). This could create a co-benefit of also reducing biodiversity impacts. However, as part of integrated global commodity markets, reductions by aquaculture producers will only help to meet emissions targets if broader food sector commitments are made. Replacing fish meal and fish oil with fishery by-products has a relatively small effect (Fig. 4b), but increased by-product utilization can improve system-wide performance when it directs potential wastes toward more favourable applications50. Finally, novel aquaculture feeds, including algal, microorganism and insect meals, are increasingly available but currently account for a small fraction of feeds. While they are likely to hold potential to improve feed quality and reduce forage fish demand51, their impacts at scale remain uncertain52 and therefore could not be modelled here.

For capture fisheries, reducing fuel use represents the primary stressor improvement opportunity. Increasing stock biomass could reduce fuel use per tonne of fish landed12,53, where a 13% catch increase with 56% of the effort54 corresponds to a 50% reduction in GHG emissions. Alternatively, we find that prioritizing low-fuel gears within each fishery can reduce GHG emissions by 4–61%, depending on the species (Supplementary Fig. 16). In some cases, this could create co-benefits for biodiversity impacts (Box 1). Another strategy is to transition fishing fleets to low-emission technologies8. While some fleets have transitioned to electric, hydrogen fuel and sail-assisted vessels, general adoption necessitates transformations beyond traditional fishery management.

Realizing blue food’s environmental potential

Blue foods already have great potential for reducing food system environmental stressors. Unfed aquaculture results in negligible values for most considered stressors, and many fed aquaculture groups outperform industrial chicken, the most efficient major terrestrial animal-source food. Capture fisheries vary widely in their GHG emissions but are low impact with respect to the other stressors considered. This underscores the value of sustainably managing wild fisheries to avoid the environmental replacement cost that would be incurred under fish catch declines24.

Our standardized estimates enhance the resolution of the potential role of blue foods within sustainable diets, highlighting opportunities to shift demand from relatively high- to low-stressor blue foods and from terrestrial animal-source foods to comparatively low-stressor blue foods. Shifting to non-animal alternatives remains an efficient lever but low-stressor blue foods may represent an appealing alternative for some consumers. Furthermore, blue foods provide the highest nutrient richness across multiple micronutrients (for example, iron and zinc), vitamins (for example, B12), and long-chain polyunsaturated fatty acids (for example, EPA and DHA) relative to terrestrial animal-source foods10, which may provide greater incentive to shift demand as consumers generally prioritize seafood freshness, food safety, health and taste over sustainability55.

Major challenges remain for shifting demand, as well as meeting increased demand. While improved management offers potential opportunities for expanding some production from low-stressor capture fisheries, uncertainty remains around the extent and feasibility of rebuilding many fisheries47. Additional research is needed to understand the total environmental impacts of large-scale expansion of low per unit stressor foods, especially for system-specific impacts (Box 1). Increasing production also requires creating appropriate incentives and reducing barriers for producers. Historical food system transitions required public investment technologies that could be scaled up by the private sector and public policy leadership56. Overly strict regulations or lack of capital can prevent expansion of low-stressor blue foods such as offshore mussel farms (see, for example, ref. 57). Facilitating low-stressor blue food expansion and novel production methods may require new and more adaptive policies and distribution of grants or other forms of start-up capital. Finally, policies can steer production and consumption through taxes and subsidies58 as well as softer policies, such as dietary advice that considers environmental impacts59.

Within the diversity of blue food production there are numerous opportunities to reduce environmental stressors. As a young and rapidly growing sector, there are many promising technological innovations in aquaculture (for example, recirculating aquaculture systems, offshore farming and novel feeds). However, less charismatic interventions may represent greater potential for rapid and substantial impact reductions. These include policy or technological interventions that improve husbandry measures (especially reducing disease and mortality) and lower FCRs. Improved management in salmon aquaculture demonstrated considerable sustainability benefits through disease and area management plans60 and improved stock management with precision aquaculture and automation61. Furthermore, selective breeding, genetic improvements and high-quality feeds can all reduce FCRs (Supplementary Table 8). While we looked at individual interventions, improvements are likely to occur through a suite of interventions and the synergistic or antagonistic interactions of interventions represents an important area for future work. Unfortunately, many innovations are often beyond the reach of smallholder producers of low-value species. This highlights a need for public research and development as well as technology transfer to enable all farmers to adopt practices that reduce environmental stressors. For capture fisheries, continued management reforms together with incentives to use low-fuel gears could substantially improve the performance of capture fisheries11,47. A range of actors will be important for stimulating a shift to more sustainable production methods and, for instance, nation states, civil society and the private sector all have important roles. Private sector pre-competitive collaborations; for example, SeaBOS62 and the Global Salmon Initiative can help to stimulate production improvements at scale. Likewise, government-led initiatives helping small-holders improve their farming practices through, for example, access to high quality feeds, seed and broodstock, are crucial for closing the aquaculture performance gap63,64,65. Certification and improvement projects can help to reduce ecosystem impacts66, but have been criticized for passive exclusion of small-scale producers. Moving towards best practices such as state-led, national certification schemes and area-based approaches will therefore be key67. Finally, the finance sector can help to steer the sector towards sustainability through strategic investments68.

The above findings do not suggest unlimited blue food growth is possible nor that expansion comes without environmental trade-offs. Furthermore, without careful consideration for local contexts and inclusion of relevant stakeholders, environmentally focused interventions can generate social and economic trade-offs that undermine broader sustainability goals. Nevertheless, farmed blue food is among the fastest growing food sectors and the global community now faces a unique window of opportunity to steer expansion towards sustainability69. Our model and results provide blue food stressor benchmarks and enable data-poor environmental stressor assessments. This serves as a critical foundation for understanding blue food environmental performance and to ensuring sustainable and healthy blue foods are available now and into the future.