Quantifying greenhouse gas emissions from global aquaculture

Global aquaculture makes an important contribution to food security directly (by increasing food availability and accessibility) and indirectly (as a driver of economic development). In order to enable sustainable expansion of aquaculture, we need to understand aquaculture’s contribution to global greenhouse gas (GHG) emissions and how it can be mitigated. This study quantifies the global GHG emissions from aquaculture (excluding the farming of aquatic plants), with a focus on using modern, commercial feed formulations for the main species groups and geographic regions. Here we show that global aquaculture accounted for approximately 0.49% of anthropogenic GHG emissions in 2017, which is similar in magnitude to the emissions from sheep production. The modest emissions reflect the low emissions intensity of aquaculture, compared to terrestrial livestock (in particular cattle, sheep and goats), which is due largely to the absence of enteric CH4 in aquaculture, combined with the high fertility and low feed conversion ratios of finfish and shellfish.

www.nature.com/scientificreports/ The geographical pattern of emissions closely mirrors production, i.e. most of the emissions arise in the regions with the greatest production: East Asia and South Asia. Emissions also correlate closely with production for most species-groups, e.g. cyprinids account for 31% of emissions and 31% of production. However, there are exceptions to this: shrimp account for 21% of emissions but only 10% of production, while bivalves produce 7% of emissions but represent 21% of production.
Production of crop feed materials (the green segments of Fig. 2) accounted for 39% of total aquaculture emissions. When the emissions arising from fishmeal production, feed blending and transport are added, feed production accounts for 57% of emissions. The bulk of the non-feed emissions arise from the nitrification and denitrification of nitrogenous compounds in the aquatic system ("aquatic N 2 O") and energy use on the fish farm (primarily for pumping water, lighting and powering vehicles).
In order to compare aquaculture emissions with those arising from meat production, the aquaculture emissions for 2010 were calculated and compared with the emissions calculated for livestock by FAO using GLEAM 4 . The FAO data 4 were used for comparison as they have the same scope and methods as the method used in this paper. 2010 was chosen as it is the most recent year for which FAO have reported global livestock results. The results of the comparison are presented in Fig. 3. This figure also includes the global emissions for capture fisheries 5 though these are for 2011 rather than 2010.
The global emissions from aquaculture are lower than livestock because (a) there is a greater amount of livestock production (in 2010 fish and shellfish accounted for 6% of global protein intake, compared to 18% of protein from meat 6 ) and (b) overall livestock has a higher emissions intensity than aquaculture.  www.nature.com/scientificreports/ emissions intensity of aquaculture. The regional average EI of each species-group is shown in Fig. 2.
For most of the finfish, the EI lies between 4 and 6 kgCO 2 e/kg CW (carcass weight, i.e. per kg of edible flesh) at the farm gate. The exception is the category "marine fishes, general", which has a significantly higher EI, due to the assumption that the ration in East Asia (and New Zealand and Australia) is 100% low value fish/trash fish (which has a higher EI than most crop feed materials) and the higher feed conversion ratio (FCR, i.e. the kg of feed input per unit of liveweight gain) of this species-group. Shrimps and prawns have high EI, due to the greater amounts of energy used in these systems (primarily for water aeration and pumping). In contrast, bivalves have the lowest EI as they have no feed emissions, relying on natural food from their environment. Within the finfish, there are some differences in the sources of GHG emissions. Species predominantly reared in Asia (i.e. Indian major carps, freshwater catfishes and cyprinids) have higher rice methane emissions, while the carnivorous salmonids have more emissions associated with fishmeal and higher crop land use change (LUC) emissions (arising from soybean production), reflecting their higher protein rations. Comparing global averages, aquaculture has a much lower EI than ruminant meat and is similar to the main monogastric commodities (pig meat and broiler meat) (Fig. 3). It should be noted that there can be significant variation in the EI of commodities, depending on factors such as genetics, feeding and farm management (for a discussion of the factors influencing the EI of ruminants and monogastrics, see 7,8 ). Fish (both finfish and shellfish) have lower EI than ruminants for three main reasons: they do not produce CH 4 via enteric fermentation, they have much higher fertility (so the "breeding overhead" is therefore much lower) and they have lower feed conversion ratios (which are a key determinant of fish EI, given the predominance of feed related emissions). Fish generally have lower FCRs than terrestrial mammals, due to the latter's higher maintenance and respiratory www.nature.com/scientificreports/ costs 9 . Being buoyant and streamlined, fish require less energy for locomotion, they are cold-blooded, and they excrete ammonia directly.

Discussion
Limitations of the analysis. The emissions are calculated for aquaculture of aquatic animals only, and therefore do not include the emissions arising from the production of aquatic plants, which constitute a significant proportion of global aquaculture production. The importance of feed is clear in Fig. 2 for all fed species. However, feed composition is constantly changing as nutritional knowledge and its application develop in response to commercial demand. This study was based on regional assumptions of feed formulations and raw material origins for the main species in the key regions. Data for this was obtained from a variety of sources (see "Methods") and updated in light of discussions with feed companies. Improved knowledge of feed formulation and raw material sourcing, combined with the overall feed efficiencies of conversion to edible seafood will help provide a more accurate picture of the overall emissions. Ultimately this would have to be done with primary data from feed companies and farmers on a case by case level.
The analyses do not include losses and emissions occurring post-farm. Depending on the specifics of the post-farm supply chain (e.g. mode of transport, distance transported, mode of processing, storage conditions), significant emissions can arise from energy use in transportation or from refrigerant leakage in cold chains 10 . However, it should be noted that all GHG emissions are attributed to the aquaculture in this study, whereas, in practice, aquaculture produces processing by-products (such as trimmings) that are often used in other sectors and the associated emissions should be allocated to these sectors.
The estimates of aquatic N 2 O should be treated with caution, as the rate at which N is converted to N 2 O in aquatic systems can vary greatly, depending on the environmental conditions. It has been noted 11 that nitrification and denitrification processes are influenced by many parameters (e.g. dissolved oxygen concentration, pH, temperature).
Finally, this study relies on data currently available in the literature. While the best available data has been used, we recommend that true empirical studies, involving primary data gathering on key parameters, should be undertaken to validate the results.
Reducing emissions from aquaculture. It has been argued 12 that because the aquaculture sector is relatively young compared with terrestrial livestock sectors, it offers great scope for technical innovation to further increase resource efficiency. They go on to identify four broad technological approaches to reducing the environmental impact of aquaculture: (1) breeding and genetics, (2) disease control, (3) nutrition and feeding, and (4) low-impact production systems. Within each of these approaches are many individual measures that could be used to reduce (or mitigate) GHG emissions.
There are many ways of reducing emissions from crop production 13 that could be employed to reduce aquaculture feed emissions. Other measures to reduce feed emissions target the efficiency of feeding. Aquaculture www.nature.com/scientificreports/ nutrition is arguably more complicated than terrestrial livestock production, in the sense that it has many more species being farmed. Each species in theory has different nutritional requirements, although the information to provide this accurately is often lacking. This drives relatively poor use of nutrients as the focus is instead on providing certain raw materials that mimic what is consumed in the wild-for example feeding high inclusions of fishmeal to some carnivorous species, in particular marine fishes. The opportunity to optimise nutrition is probably greater in aquaculture than in terrestrial species, since much greater research effort has been focussed on terrestrial species to date. Some mitigation measures may be quite expensive while others are relatively cheap or may even reduce costs. In order to achieve the twin goals of reducing emissions, while increasing the supply of affordable protein, we need to analyse the effects that introducing measures may have on farm profits and emissions. Cost-effectiveness analysis (CEA) can help us to understand these effects.

conclusions
Aquaculture is a biologically efficient way of producing animal protein compared to terrestrial livestock (particularly ruminants) due largely to the high fertility and low feed conversion ratios of fish. The biological efficiency is reflected in the relatively low prices and emissions intensities of many aquaculture commodities. However, the moderate GHG emissions from aquaculture should not be grounds for complacency. Aquaculture production is increasing rapidly, and emissions arising post-farm, which are not included in this study, could increase the emissions intensity of some supply chains significantly. Furthermore, aquaculture can have important non-GHG impacts on, for example, water quality and marine biodiversity. It is therefore important to continue to improve the efficiency of global aquaculture to offset increases in production so that it can continue to make an important contribution to food security. Fortunately, the relatively immature nature of the sector (compared to agriculture) means that there is great scope to improve resource efficiency through technical innovation, often in ways that reduce emissions while improving profitability. CEA can be used to help identify the most cost-effective efficiency improvements, thereby supporting the sustainable development of aquaculture.

Scope.
The system boundary is "cradle to farm-gate". It is recognised that significant emissions (and losses of product) can occur post-farm during transport, processing and distribution.
Global aquaculture is a complex sector consisting of many different species reared in a variety of systems and environments. In order to manage this complexity, the analysis focuses on the main cultured aquatic animal species-groups (aquatic plants are excluded), i.e.: bivalves, catfish, cyprinids, freshwater fish (general), Indian major carps, marine fish (general), salmonids, shrimps and prawns and tilapias. The main species-groups were identified by extracting production data from FAO 14 , listing the species-groups within each geographical region (according to FAO definitions) in order of production amount, then selecting the groups until they accounted for > 90% of the production within the region (> 85% in Eastern Europe). This approach captured an estimated 93% of global production. The categories of GHG included in the analysis are summarised in Table 3. carbon sequestration in pond sediments. Carbon sequestration in pond sediments is not included in this study. While it has been suggested 15,16 that ponds could act as net carbon sinks, questions have been raised about the certainty of the abatement rate and permanence of the sequestration 17 . There are also concerns about impacts on water quality and fish health arising from the nutrient inputs associated with carbon sequestration 18 . calculation method. The method is summarised in Fig. 4 and further details are provided below. Table 3. Summary of the GHG categories included in the calculations.

Name Description
Feed: fertilizer production Emissions arising from the production of synthetic fertilizers applied to crops based on the values derived using GLEAM 4 . Regional average values were used for each feed, meaning that the EFs at least partially capture variation in crop production efficiency between regions. EFs for additional feeds (e.g. fishmeal, poultry meal, feather meal, meat & bone meal, blood meal, groundnut meal) were derived from Feedprint 19 and EFs for fish oil from 20 . Non-commercial feed materials were assumed to be produced locally, and have different emission profiles to their commercial equivalents (e.g. no emissions from transport). feed conversion ratios and ration composition. A distinction was made between two types of aquafeed as follows: (a) commercial aquafeed, which are compound feeds purchased from specialised feed manufacturers and/or feed wholesalers/retailers. The feed is comprised of materials sourced nationally and internationally, which are formulated and blended into high quality compounded pellet feeds and (b) farm-made/ semi-commercial aquafeeds (which often include mashes or wet pellets) made on the farm or produced by small-scale feed manufacturers from locally sourced feed materials. The proportions of production reared on commercial and non-commercial rations were estimated based on 22 . Changes in commercial conditions make it difficult to keep up to date through academic papers, as the feed compositions are improved/changed frequently and farming conditions fluctuate with improvements and emerging disease challenges. To account for this, feed composition (protein and energy), raw material rations and economic feed conversion ratios (eFCRs, which take into account average mortalities) were derived from a range of sources including: AFFRIS 23 , FAO publications [24][25][26] , journal articles (e.g. 22,27 ), grey literature (e.g. 28 ) and expert opinion to reflect the most recent updates. Feed conversion ratios used and the ration assumptions for the main culture group x location combinations are given in the Supplementary Information. total production by species-group and region. Production data for 2017 was extracted from the FAO database FishStatJ 14 .
on-farm energy use. Energy is used on fish farms for a variety of purposes, primarily for pumping water, lighting and powering vehicles. The average amount of energy required to produce one tonne of live weight of fish and shellfish, and the proportions of electricity, diesel and petrol used, were calculated based on values presented in the literature (see Supplementary Information). The rates of electricity, diesel and petrol used per tonne of live weight (LW) were then multiplied by emission factors to determine the emission intensity (see Supplementary Information). Global EFs were used for petrol and diesel, and regional EFs were used for grid electricity 29 .