Seafood more nutritious at lower emissions than terrestrial animal protein sources
At the highest level, we find that while a two-fold variability exists between the average nutrient density scores of major seafood species groups, median GHG emissions vary by over an order of magnitude between sources of seafood—both in terms of species groups but also in terms of how seafood is produced (Fig. 1). The global average performance of all seafood assessed, weighted by species production volume (the two lines in Fig. 1), has a higher nutrient density than beef, pork and chicken and lower GHG emissions than beef and pork. Importantly, the weighted average performance of all seafood species analysed is useful only for comparative purposes—as it does not per se indicate whether this average represents a high or low value. This is true also for the nutrition and environmental data shown in a recent study28, whose relative values do not reveal whether the difference between the best and the worst-performer is small or large.
Relative nutrient density and production-related GHG emissions (i.e. post-harvest emissions are excluded) per edible weight of globally important seafood groups from fisheries (solid colours) and aquaculture (striped) at the point of landing or harvest, along with beef (B), chicken (C) and pork (P). Relative size of seafood group bubbles is proportionate to 2015 global edible weight production volumes, and GHG and nutrient density values are weighted by species. Both GHG values and nutrition scores are log-weighted and positioned relative to the weighted average of all included seafood species (3.7 kg CO2e per kg edible weight and NDS 4.0). For graphs showing species in each species group individually, see Supplementary Figs. 1–8.
Pelagics, wild salmonids and farmed bivalves best performers
Amongst the seafood groups defined, wild-caught salmonids (pink and sockeye salmon) and the small pelagic species (e.g. herrings, mackerels, and anchovies) and farmed bivalves have the lowest GHG emissions per nutrient density ratio (Fig. 1, Table 1), and comprise the top tertile of species considered (Table 1). These are not the most consumed seafood species, though. Harvest of wild salmonids is relatively low and functionally constrained by limited stocks. A large portion of landings from many small pelagic fisheries is currently destined for other uses (e.g. inputs to aquaculture and livestock feeds), largely due to insufficient demand for direct human consumption, but also as a result of incentives created by regulations. Farmed bivalves (e.g. oysters, mussels etc.) are found among top performers in terms of GHG emissions, but provide slightly lower nutrition density. In contrast, crustaceans, both farmed (primarily tropical shrimp species), and wild-caught (various shrimp species, American lobster, etc) and cephalopods all result in higher than average emissions while providing lower than average nutritional scores. Our findings with regard to best- and worst performing species and species groups confirm previous findings26,27,28. Unlike other seafood groups that are defined phylogenetically and presumably are nutritionally more similar9 (see also Fig. 1), species grouped as ‘whitefish’ simply share desirable human consumption characteristics (e.g. firm, pale flesh, mild flavour). As both wild-caught and farmed whitefish represent substantial production tonnages, their group-specific nutritional and GHG emission scores have a commensurate influence on the overall weighted average scores for all seafood products analysed. Despite this, the wild-caught whitefish species considered resulted in the lowest nutrient density scores of all groups assessed (Fig. 1). In all comparisons between species groups and production forms in Fig. 1, it is important to keep in mind that as the observations within each group and methods used to characterize attributes are not fully consistent across data sources, it was not considered feasible to conduct formal statistical testing. The differences observed should therefore be interpreted as indicative based on currently available data.
Large variability within species groups
Each species group consists of 2-10 species with considerable variability within each group either in terms of nutrient density scores, production-related GHG emissions, or both, with larger variation observed in GHG emissions (Fig. 2, Table 1, see also Supplementary Figs. 1–8). The lower variability in nutrient density is partly due to the choice of capping nutrient content at the dietary reference intake (DRI, i.e. nutrient content exceeding the DRI does not influence the nutrition score, see Methods). This choice also reduces the contribution of nutrition to the combined score, which is driven moreso by GHG emissions due to the greater variability in emission rates. Likely reflecting the large diversity of species and production sources contained within the whitefish group, it encompasses the widest species-specific variability of both GHG emissions and nutritional density (Fig. 2).
Nutrient density scores are based on the 21 nutrients common to all species (full bars) and, where possible, 23 nutrients (grey lines) (for nutrients see Methods). GHG emissions of individual seafood species are representative of the dominant production method for each (or weighted if multiple major production method is employed globally). Solid bars indicate species from fisheries, and striped bars species from aquaculture. Comparisons to land-based animal proteins are based on nutritional content of averaged meat cuts for beef and pork, and fillets for chicken. GHG emissions of beef are beyond the scale at 56 kg CO2e per kg edible product.
Within most species groups, the large observed variability in GHG emissions (Fig. 2) mainly determined by the production technique, suggests the potential for substantial reduction opportunities related to production methods and practices. Small pelagics are an exception and exhibit greater variation in nutrient density values than in GHG emissions. While the nutritional content of species can only be influenced to a limited degree (e.g., through changes in feed composition, timing of harvest), substantial opportunities to reduce GHG emissions exist by increasing use of low-emission technologies or inputs (e.g. energy sources, feed inputs), and by rebuilding stocks30,31. In particular, changes in feed composition and use represent major opportunities to reduce emissions from fed aquaculture32. Greenhouse gas emission intensity values reported in Fig. 2 are also influenced by differences in edible yield, which is high for cephalopods (~70–80% from liveweight), but low for bivalves (~15–25% from live weight), with finfish species falling between these extremes. Yields can be maximized through technical or temporal regulations or through improved processing technology, but each species has biological limits. Analysing these parameters on a species group level28 misses out on this variability, which can be larger than the between-group variability26.
Half of seafood species perform better than terrestrial animal protein sources
When compared to land-based animal source foods, 22 of the 41 seafood species whose nutrient density scores could be assessed (54%) and 17 of the 34 seafood species whose production-related GHG emissions could be quantified (50%) performed better than beef, pork and chicken (Fig. 2). Pork performs just below the average of seafood in both dimensions assessed, whereas chicken has a much lower nutrient density, comparable to the poorest performing seafood groups (Fig. 1). Beef scores just under the average seafood nutrient density but results in higher GHG emissions than any other food analysed here (Figs. 1, 2). Nutrient density of seafoods is only lower than land-based animal products for a few species (e.g. some whitefish and Japanese carpet shell; Fig. 2).
Nutrient density of seafood is driven by various nutrients
Our results indicate that salmonids, both wild-caught and farmed, and small pelagics are the most nutrient dense seafoods assessed (Fig. 2, Table 1), which is consistent with previously reported research26,27. Some tuna species and common carp also rank high for NDS21 (tertile 1, Table 1). Most bivalves and cephalopods showed intermediate nutrient density scores (tertile 2, Table 1), whilst most crustaceans and whitefish species ranked low (tertile 3, Table 1).
Vitamin B12, niacin and vitamin D are the nutrients that, on average, made the greatest contribution to nutrient density scores across the seafood species assessed (on average 20, 12 and 9% of NDS21, respectively; Fig. 3). Vitamin B12, in particular, had the greatest impact on nutrient density in 35 of the 41 species assessed for NDS21 (see Supplementary Table 1). Analysing the nutritional value at an overall taxonomic level risks missing important within-group differences and average values may not represent individual species or the group well. For example, our data showed that oysters had the highest calcium content of all species included and that all bivalves were over average in calcium content, whereas a recent study28 concludes that bivalves contain less than average calcium across species groups. Despite differences among individual species, patterns can be identified between seafood groups in terms of nutrient composition (Fig. 3). For example, n-3 fatty acids are the key contributor to NDS21 amongst the nutrient dense salmonids (especially when farmed) and small pelagics (Fig. 3). Farmed whitefish and wild salmonids are particularly rich in vitamin D, whereas cephalopods, mussels and crustaceans (both farmed and fished), have a high copper content (Fig. 3). Partly due to the higher vitamin D content, farmed whitefish performed better than most of fished whitefish species assessed. Due to the role whitefish occupies in current consumption, this aspect would probably deserve further investigation by broadening the analysis to include a larger number of farmed whitefish species (e.g. more carp species).
Nutrient profiles of seafood groups expressed as percentage contribution to the nutrient density score NDS21, calculated as weighted averages of included species within groups based on edible production volumes. Desirable nutrients are visualized if contributing ≥10% of nutrient density for at least one group. Other includes all remaining desirable nutrients not reaching the cutoff value.
Nutrient density is a composite index to which desirable and undesirable nutrients contribute positively and negatively to the final score. Our analysis confirms that seafood is a source of animal protein that delivers minimal quantities of sodium and saturated fat, nutrients that are indeed considered ‘undesirable’ from a public health perspective. Bivalves are the only group considered in which sodium can be regarded as non-negligible (Fig. 3).
The importance of individual nutrients for adequate nutritional intake and overall health differs between populations and population groups. To account for such differences it has been suggested that nutrient density scores should be tailored to the targeted population assessed23. Here we intentionally avoid making such local adjustments in order to be able to describe overall patterns of climate impact in relation to nutrient content across species. Results therefore should necessarily be considered within this context and goal in mind. Any future applications of the approach to specific populations should attempt to account for the dietary needs locally or of specific population sub-groups (defined by age, gender or socioeconomic parameters), as well as local availability of seafood products and their source.
The nutrient profiles of many species analysed revealed very high concentrations of a few nutrients, often well above their DRI. As described, capping was applied in the NDS calculation even though this flattened the nutritional variability among species. If not capped to 100% of DRI, disproportionately high scores would be seen even in products providing low levels of most nutrients and high levels of one or a few nutrients (e.g. Japanese carpet shell). Vitamin B12 was the nutrient that most often exceeded the DRI, in 26 of the 41 seafood species assessed, with contents up to almost 25 times the DRI in the flesh of some species (see Supplementary Table 2 for data and Supplementary Discussion 1).
In dietary guidelines, vitamin D, n-3 fatty acids, selenium and iodine are often identified as nutrients from seafood of special importance in human diets2. Nutrient profiles are highly variable between species (Fig. 4) and two types of seafood with the same nutrient density score can make markedly different contributions to the intake of specific nutrients. By assessing the role of individual nutrients beyond their contribution to the nutrient scores, potentially important sources of specific nutrients can be identified despite their relatively small concentrations (Fig. 4). Largehead hairtail, for example, only displays an intermediate nutrient density score (Table 1) but is the richest source of n-3 fatty acids among the analysed species (Fig. 4, Supplementary Table 2). Common carp and Nile tilapia are farmed whitefish species with markedly different nutrient densities (Fig. 2, Table 1, Supplementary Table 1), but both containing relatively high concentrations of vitamin D (Fig. 4, Supplementary Table 2). Composition data on selenium and iodine were only available for a subset of the studied species (36 out of 41). Amongst these, tunas were all excellent sources of selenium, but even some crustaceans like Gazami crab exceed the DRI for this element (Fig. 4). Iodine is a nutrient that many humans are deficient in globally33. Some species, like American lobster, Atlantic cod and haddock, are good sources of this micronutrient (Fig. 4), despite not scoring well in terms of NDS21 (Fig. 2, Table 1). With the exception of these species, the addition of selenium and iodine to an NDS23 did not markedly affect the overall patterns of nutritional performance of the species considered (see NDS23 in Fig. 2 and Supplementary Table 3). In order to capture the full potential of seafood species in the human diet, it is, however, advisable to include these minerals in nutrition evaluations when their concentrations are available. Additionally, it is important to be aware that the NDS only relies on the nutrient content as a measure of nutritional quality. Other aspects related to the potential health effect of consuming seafood, such as nutrient bioavailability, food matrix effects, content of other bioactive or toxic compounds are not captured by this method.
Nutrient profiles of six seafood species as percentage contribution of 100 g raw edible flesh to the dietary reference intake: Atlantic cod (A), Common carp (B), Albacore (C), Largehead hairtail (D), American lobster (E), and Gazami crab (F). The twenty-one nutrients included in the NDS21 formula are represented here, plus selenium and iodine (not available for Common carp or Largehead hairtail; iodine not available for Gazami crab). Nutrients are grouped by proteins and fats (orange), minerals (blue), and vitamins (green). Values are displayed as relative area of the pie slice, with a maximum value (full slice) representing a 100% contribution. All nutrients with the exception of n-3 are capped to 100% when exceeding the DRI.
Our results are broadly consistent with those of similar analyses undertaken for seafood consumed in Sweden26 despite using distinct data sources, with pelagic and salmonid species performing best. Where differences exist (e.g. the ranking of oysters or lobster), they result from differences in methodological choices made for calculating the nutrient density (e.g. capping of nutrients and a different selection of nutrients), and reliance on a wider suite of nutrient composition data sources. Koehn et al.27 also identified small pelagics, and salmon as best-performers, despite large differences in the modelling of the nutrient index, e.g. excluding the nutrients that were most important in this analysis (Vitamins B12, D and niacin) as well as content of undesirable nutrients, calculating the average instead of the sum of nutrient content per DRI ratios and letting content higher than DRI influence the index for all nutrients.
Nutrient content assessment methods and reporting for seafood species varies widely both within and between nutrition databases and likely affected the resulting nutrient density scoring and combined nutrition-climate impact assessment. Relatedly, nutrient densities for each species are calculated based on single observations, rather than averages from multiple databases29. Additionally, many globally important species (e.g., carp) could not be included due to lack of detailed nutrient composition which points to important data gaps. This suggests the need to include more seafood products in methodologically harmonised, public food composition databases.
Pelagics have lowest greenhouse gas emissions, crustaceans highest
Of the 41 seafood species for which an NDS21 value could be assessed, we were able to quantify production mode specific (e.g. fished or farmed) GHG emissions for 34 species (Supplementary Table 4). Emission intensities of individual species and major patterns of relative emissions associated with seafood groups (Fig. 2) are broadly consistent with previous findings in the study of Swedish consumption26. Differences occur when globally important production technologies, modelled here, differ markedly from the specific sources known to supply Swedish consumption. Moreover, emission data that were not available at the time of the previous analysis were used for a few species (e.g. oyster and Atlantic salmon).
Wild-caught crustaceans and some farmed whitefish, tunas, farmed salmonids and cephalopods had the highest GHG emission intensities. Both farmed bivalves assessed (mussels and Pacific cupped oysters) together with all eight small pelagic species assessed, pink and sockeye salmon and Alaska pollock, all had emission intensities far below that of chicken (Fig. 2), while no seafood species approached the scale of GHG emissions from beef.
Five of the seven species for which a GHG emission value could not be identified or characterised directly are farmed bivalve species (Jackknife clam, Japanese carpet shell, farmed scallop, green and Chilean mussels) which points to a major gap in the LCA and related GHG emission accounting literature. This is unfortunate given the promising performance of bivalve species described previously in terms of emissions11,32 and nutrition26,28. As GHG emission values were available for blue mussel culture, it was used to characterise cultured green mussels and Chilean mussels, as a reasonable first approximation of their actual emission intensities. The five species (four farmed and one fished) that could not be assigned a GHG emission value are all produced primarily in China (Fig. 2).
Fuel and feed dominate seafood emissions
Fuel combustion during fishing is the primary source of GHG emissions from capture fisheries, with fuel use intensity (FUI) rates strongly influenced by the fishing gear employed and the relative abundance and catchability of stocks30,31,34,35. Consequently, amongst the fished species assessed here, those with higher emission intensities were typically landed using more fuel-intensive fishing methods or targeted species that are less abundant (Supplementary Table 5). For example, amongst the four tuna species assessed, those for which a larger proportion of total landings are caught using hook and line gears (bigeye, albacore), have higher emission intensities than those species primarily caught using purse seines (yellowfin, skipjack). The case of Alaska pollock presents an interesting example of a species in the whitefish group that performs very well due to a relatively fuel-efficient fishing method, pelagic trawling, resulting in a high catch rate and a remarkably low emission intensity, confirmed by very recent data36.
Sources of GHG emissions from aquaculture production are far more diverse37,38 though tend to be higher when species in culture are fed, especially if also substantial energy inputs are required to maintain culture water quality (e.g. aeration, waste removal, chilling, etc). Amongst the farmed species assessed, Amur catfish had the highest GHG emission intensity while farmed mussels and oysters, both unfed in culture, had the lowest (Supplementary Table 4). Relatively low edible yield rates are a secondary explanatory factor behind the relatively high emissions from crustacean and catfish production.
Nutrition a more relevant basis for comparison than liveweight
Communicating the environmental performance of seafood products based on their nutrient density more completely captures the function of these products relative to the performance of the systems providing them. It improves upon comparisons made on the basis of edible weight, which in turn, is a substantial improvement over comparisons made on liveweight. Importantly, this nutritional lens not only facilitates comparisons between species using a more product-relevant basis, but substantially changes the result of those comparisons. In fact, when scaling up results to annual global production volumes of the species and species groups that we have analysed here, the most important groups in terms of total liveweight mass produced are, in descending order: bivalves (oysters and mussels), farmed whitefish species, small pelagics and wild-caught whitefish species (Fig. 5). Wild-caught salmonids, crustaceans and cephalopods represent the smallest liveweight tonnage species groups produced (Fig. 5). On the basis of mass of edible product available, small pelagics and farmed whitefish dominate. This is because small pelagic species have higher edible yield rates (53–62%, Supplementary Table 6) than most other groups but in particular compared to farmed whitefish species (37–45%) and bivalves with the lowest edible yields of all groups (15–24%). Multiplying edible mass of each species group, calculated using species-specific edible yield factors, by the weighted average NDS21 score of species within each group, the importance of small pelagics in terms of potential human nutrition increases further (Fig. 5). Other groups whose relative nutritional importance to humans increases when moving from volumes to nutrition density are large pelagics, farmed whitefish and salmon, while the relative importance of fished whitefish, cephalopods, farmed and wild-caught crustaceans, and bivalve groups are all reduced from their contributions to total edible volume. When translated into GHG emissions, the differences are even more pronounced with the three top-performing species groups, small pelagics, wild salmonids and bivalves, together representing 35% of the available nutrition density while only contributing 6% of production-related GHG emissions across all species assessed. In contrast, farmed and fished crustacean species represent 8% of total seafood nutrition density and produce 17% of total emissions estimated across all species assessed. All data used to produce graphics is presented in Supplementary Data 1 and Supplementary Material.
Contribution of seafood groups to global production volumes in 2015 expressed on a live and edible weight basis (for species-specific edible yields see Supplementary Table 6), nutrient density (NDS21) and GHG emissions, all weighted by species within each seafood group. Solid groups consist of wild-caught, striped ones of farmed species.
In addition to strategies to improve the nutritional output from individual seafood systems noted above (i.e. changes in feed, timing of harvest), larger-scale opportunities exist to increase the nutritional performance of seafood systems more broadly. Policies and technological innovation that increase direct consumption of landings from small pelagic fisheries could result in dramatic improvements in the nutritional output of global fisheries while limiting emissions. Although the proportion of global seafood production destined for non-food purposes is declining1, in many settings utilisation of small pelagic species for feeds is still incentivised. For example, policies aiming to concentrate fisheries to fewer, larger vessels landing larger volumes result in reduced catch quality, and a larger proportion of catches ending up as feed, also because of limited capacity to process these landings before quality deteriorates further.
Policy changes that facilitate greater utilisation of landings for food could take many forms (e.g. quota reallocations, distributing harvest opportunities in time or space, improving on-board and in storage product conservation, differential resource rents based on product destiny, etc.), but will need to take into account the unique characteristics of individual fisheries and their settings. Moreover, for any substantial change to be successful, many actors need to be involved beyond fish harvesters, including the food industry and retailers who are going to need to develop, produce and sell new products. Efforts will be needed to understand consumer attitudes towards these species whose top performance in both dimensions, nutrition and climate, would call for using a much higher proportion directly as food than is the case today. In addition, food product innovation designed to increase the utilization of fish or by-products from fish in supplements could contribute to making seafood more accessible to consumers in both high- and low-income country settings1,39. Separately, policies that facilitate the expansion of mussel farming, together with efforts to increase mussel consumption (e.g. dietary advice recommending mussels, sponsored cultural events featuring mussels, developing convenient and affordable mussel-based food products etc.) would also improve the combined nutritional and climate impacts of seafood consumption in general. While macroalgae species were not included here, a previous analysis of environmental stressors did32 and found that seaweed was a promising low-impact group of species. There are big knowledge gaps related to the content and bioavailability of nutrients from seaweeds, as well as about their content of undesirable substances, but research is ongoing on these topics. Species from unfed, low-trophic aquaculture have been identified to have a large potential as future foods10,29,40.
Conversely, just as dietary advice in many countries recommends against intake of red and processed meat, advice related to seafood consumption could indicate types to avoid based on lowest nutritional value at highest emissions. For example, the European Commission is, as part of its Green Deal policy, developing nutrition and sustainability labelling for food products in the coming years and as more data become available this type of analysis will become easier, more robust and informative in that kind of effort. From a global perspective, it may even be wise to promote the most nutritious forms of seafood in nutrient-deficient populations and communities, even when production results in relatively higher emissions, while in populations not at risk for nutrient deficiencies, consumers could give more attention to the emissions than to nutritional content when choosing seafood products for their diets. In fact, it is in nutrient-deficient population groups that any increase in seafood consumption would have the most positive effects for human nutrition. Our results show that nutrition-based functional units can be a valuable complementary tool when comparing the environmental performance across seafood species and other foods.
Seafood statistics and research often defaults to a production perspective, with even consumption being measured in liveweight mass1. If human nutrition is the ultimate objective of fisheries and aquaculture, it is important that outputs are understood and evaluated on a nutritionally relevant basis particularly given the diversity of species involved4 and maximising the nutritional output while minimising environmental costs of seafood provisioning should be a guiding principle for policy-making in these areas41. As both fisheries and aquaculture face many environmental challenges—in terms of sustainable utilisation of stocks, reduction of by-catch and impacts on local ecosystem structure and function, nutrient enrichment, and disease amplification—restricting the analysis of sustainability to GHG emissions may seem very limited. However, it is not uncommon to see biotic impacts aligning with climate impacts as carbon-intensive fishing methods often also result in larger ecosystem impacts30,31,42,43,44. In such cases, relative rates of GHG emissions can serve as a rough indicator of broader environmental sustainability, although there are important exceptions when GHG emissions and broader environmental impacts do not align, e.g. when comparing open and closed aquaculture systems45. Ideally, nutrition data and key emissions drivers, fuel use in fisheries and feed use and composition would be collected and made available in a standardised way to facilitate and increase the robustness of this type of synthetic analysis and comparisons across species, species groups and production technologies. This would also enable monitoring of performance over time, which could help guide us into a future of nutritious foods at low environmental costs.
