Resource Use and Pollution Potential in Feed-Based Aquaculture

Abstract

Feed use in aquaculture results in large amounts of embodied land, freshwater, energy and wild fish use. Selection of feed ingredients at feed mills can reduce the amounts of one or more of the four major natural resources embodied in feed. However, better feed management to lessen FCR is more likely the key to lessening resource use at the farm level. Of course, lessening the FCR will reduce the amount of feed that must be purchased and diminish the direct and embodied negative environmental impacts associated with feed. It also is important to note that mechanical aeration applied in many methods of production requires more energy than associated with feed alone. Aeration is necessary for high feed inputs required in intensive production, and without aeration, most types of intensive production would not be possible. The amounts of resource use attributed to feeding and aeration were applied in estimating the resulting quantities of water pollutants in effluents and emission of atmospheric contaminants. Some of the misunderstandings about life cycle assessment (LCA) such as it usually covering all impacts of product systems, and especially its failure to assess the oxygen demand of effluents are mentioned.

Keywords:

• Resource use for feeds
• feed-based aquaculture
• fishmeal and oil
• feeds and environment
• feed ingredients

Introduction

Studies of farmed shrimp production have revealed that increasing intensification through the use of feeds results in progressively less land and freshwater use per tonne of shrimp production. More energy is required for producing 1 tonne of shrimp in feed-based aquaculture than is necessary in extensive culture without feed input. The increase in production possible per each equal increment of feed and aeration applied is rather constant for a given type of culture, and energy use per tonne of shrimp production does not increase with greater intensification of feed-based shrimp culture in aerated ponds (Boyd et al. 2021a). The same principle should be applicable to fish production in ponds and possibly other culture systems (Davis et al. 2021a, 2021b).

Intensive feed-based aquaculture in ponds and most other culture systems besides raceways and cages requires not only large inputs of feed but considerable quantities of electricity and hydrocarbon fuels to power aerators. Liming materials may be required to restore alkalinity depletion by acidity resulting from the biological oxidation of ammonia excreted by the culture animals. Molasses, raw sugar, or other carbohydrate source may be applied, especially in biofloc technology, to enhance the conversion of ammonia to microbial protein edible by shrimp and some fish.

The inputs of feed into aquaculture systems may have direct negative environmental impacts such as discharge of feeding wastes in pond effluents and onsite emissions of greenhouse gases (GHGs) by hydrocarbon fuel use, neutralization of acidity with liming materials, fish respiration, and decomposition of feeding waste within culture systems. In addition, what are known as embodied resource use and embodied, negative environmental impacts result from the extraction, fishing, or agricultural production of raw materials and by processing, production, and delivery of the products for end use in aquaculture (Boyd et al. 2021a; Boyd and McNevin 2015, 2021; Chatvijitkul et al. 2017a).

This assessment was done for several reasons. The original resource use coefficients for feed ingredients derived by Chatvijitkul et al. (2017b) were revised based on new information related to the production methodologies for the ingredients. The original estimates of embodied land, freshwater, and energy in aquaculture feeds were based mainly on experimental formulations from the literature. In the present effort, typical, generic, commercial feed formulations will be used to calculate embodied land, freshwater, and energy in feeds for several species. The new information on feeds will be used to show that life cycle assessment (LCA) is not the only way of estimating the potential impacts of feeds.

Effects of inputs to aquaculture cannot be prevented at the farm level, but efforts at the farm level can reduce the amounts of pollutants released and the embodied effects associated with aquaculture production. This is especially true for feeds; reducing the feed conversion ratio (FCR) from 1.6 to 1.5 results in 100 kg less feed input per tonne of production. More effective use of feeds will lower the expenditure for feed, but it also will eliminate the embodied effects associated with the amount of feed spared through more efficient conversion to harvest biomass. A lower ratio of feed applied to biomass harvested also lessens the amounts of nitrogen and phosphorus, the plant nutrients usually responsible for eutrophication, released into culture systems in waste (Chatvijitkul et al. 2017a). In addition, at lower FCR, less waste organic matter from feed will be made available for microbial oxidation, and this will diminish the amount of aeration necessary per tonne of production and lower energy use and expense (Boyd and McNevin 2021). Environmental benefits result from less waste production, because less waste diminishes aerator use and carbon dioxide emissions incurred through energy use by aerators and also from oxidation of organic waste. Careful attention to resource use can have positive effects to the economics of production, resource conservation, and environmental protection.

Feed ingredient coefficients

Feed is widely used in aquaculture; about 73% of the production of farmed fish and crustaceans is feed-based, and in 2017, a calculated 51.2 million tonnes of feed were consumed (Naylor et al. 2021). An animal feed production survey reported that 1.27 billion tonnes of compounded animal feed were produced worldwide in 2022 (Alltech 2023) of which 52.9 million tonnes were for fish and shrimp. Data of Tacon et al. (2022) suggest that feed consumption was already 52.7 million tonnes in 2018. While aquaculture feeds comprise only around 4 or 5% of global, compounded animal feed production, the greatest shares of the annual, global fishmeal and fish oil supplies become ingredients for aquaculture feeds (Boyd et al. 2022; Naylor et al. 2021). Fortunately, there is evidence that the aquaculture feed industry is reducing the inclusion rates of these natural resources of limited supply (Naylor et al. 2021; Tacon and Metian 2008; Tacon et al. 2022). The reliability of resource use coefficients for fishmeal, fish oil, and animal processing byproduct meals and vegetable oils often used to replace fishmeal and fish oil in aquaculture feed production are particularly important in efforts to assess resource use for aquaculture feeds. The coefficients for these feed ingredients used in earlier computations of resource use in aquaculture feeds (Chatvijitkul et al. 2017b; Boyd and McNevin 2022) were based mainly on data presented by Davulis and Frick (1977).

Data from Peruvian fishmeal plants (Fréon et al. 2017) suggests that energy use was about 6.33 GJ/t product, but fish oil is not mentioned separately and the energy use for trawling was not included. Hilmarsdóttir et al. (2022) studied fishmeal plants in Iceland and reported that trawling energy per tonne of fishmeal was 4.20 GJ/t and processing plant operation and clean up required an average of 8.43 GJ/t [somewhat greater than reported by Fréon et al. (2017)] for a total of 12.63 GJ/t fishmeal. The yield of fish oil per tonne of fishmeal produced varied with the pelagic species used as raw material as follows: caplin, 187 kg; a mixture of mackerel and herring, 850 kg; blue whiting, 28 kg. The fishmeal yield from raw material was rather consistent at 16–20%. Because of the great variation in fish oil yield of 0.5–17%, it is best to rely on the energy use data for plant operation with the average annual production of fishmeal and oil from raw material of about 212 kg oil for each tonne of fishmeal produced (Boyd et al. 2022) in allocating the resources used in producing fishmeal and fish oil from whole pelagic fish on a weight basis. The sum of coproducts is 1.212 t of which 83.3% is fishmeal and 16.7% is fish oil. Based on these percentages, energy use by weight of coproduct is 10.52 GJ/t for fishmeal and 2.11 GJ/t for fish oil.

Fishmeal also is produced from fish processing waste, and the yield is about 26.5% fishmeal and 4.2% fish oil, and each tonne of fishmeal can be equated to 158 kg fish oil (Boyd et al. 2022). Allocating the rendered fishmeal and fish oil results in 86.3% fishmeal and 13.7% fish oil, but the energy use would not include trawling energy. The allocation indicates 7.28 GJ/t for rendered fishmeal and 1.15 GJ/t for rendered fish oil. These estimates of energy use for both fishmeal and fish oil from wild fish and rendered fish waste are considerably less than the values of 18.60 GJ/t from Davulis and Frick (1977) as reported by Chatvijitkul et al. (2017b).

Land and freshwater use for fishmeal production was given as 55 m²/t and 87 m³/t, respectively (Hilmarsdóttir et al. 2022). Allocating between the two coproducts gives 45 m²/t and 72 m³/t for fishmeal and 10 m²/t and 15 m³/t for fish oil.

Energy use for processing slaughterhouse waste into byproduct animal meal (UK class 1) was estimated to be 3.91 GJ/t (Ramirez et al. 2012). This also is considerably less than the value of 8.60 GJ/t reported by Chatvijitkul et al. (2017b).

The recalculations of resource use coefficients related to marine fishmeal and fish oil and to both aquatic and terrestrial meat animal processing byproduct meals and oil along with some additions and recalculations of plant-source feed ingredients were used to update the original calculations of embodied land, freshwater, and energy use coefficients presented by Chatvijitkul et al. (2017b). These updated coefficients are included in the revised lists of coefficients (Tables 1–3).

Table 1. Revised embodied land, water, and energy use coefficients for animal products commonly included in aquaculture feeds.^a

Feedstuff	Crude protein (%)	Land (ha/t)	Water (m^3/t)	Energy (GJ/t)
Blood meal	76.7	0.100	2500	3.91
Bone meal	26.0	0.250	2500	3.91
Bone and meat meal	52.8	0.250	2500	3.91
Fat and oil, animal	–	0.100	2500	3.91
Feather meal	85	0.06	100	3.91
Fishmeal, marine	65	0.00055	87	10.52
Fishmeal, rendered	50.5	0.0006	120	7.28
Fish oil, marine	–	0.00010	15	2.11
Fish oil, rendered	–	0.0002	30	1.15
Fish solubles	41.7	0.0006	120
Pig byproduct meal	57.0	0.300	3006	3.91
Poultry byproduct meal	55.0	0.200	2069	3.91
Shrimp head meal	53.0	0.0006	100	3.91
Shrimp waste meal	46.7	0.0006	120	7.28
Squid meal	77.7	0.00019	100	10.52

^aRevision of tables presented by Chatvijitkul et al. (2017b) and Boyd and McNevin (2022) based on data from fishmeal and fish oil production presented by Fréon et al. (2017) and Hilmarsdóttir et al. (2022) and slaughterhouse operation by Ramirez et al. (2012).

Table 2. Revised embodied land, freshwater, and energy use coefficients for plant-based products commonly included in aquaculture feeds.^a

Ingredient	Crude protein (%)	Land (ha/t)	Freshwater (m^3/t)	Energy (GJ/t)
Barley grain	11.9	0.341	1423	2.70
Beans (faba)	30.0	0.307	2018	4.25
Brewer’s dried grain	25.0	–	100	11.87
Canola seed (rapeseed)	18.6	0.505	2271	4.94
Canola meal	35.6	0.288	1115	8.05
Canola oil	–	0.217	4301	6.38
Cassava tubers	3.5	0.100	550	0.86
Cassava meal	0.9	0.512	1878	6.21
Cassava starch	2.5	0.363	1052	5.53
Coconut meal	21.0	0.736	2093	2.77
Coconut oil	–	0.273	4490	1.05
Corn grain	9.5	0.171	1222	3.64
Corn gluten feed	21.5	0.031	1202	1.76
Corn gluten meal	60.2	0.011	1202	0.61
Corn meal	9.0	0.250	1081	4.80^b
Corn oil	–	0.001	2587	3.65
Corn starch	0.3	0.114	1671	6.37
Cottonseed meal	41.4	0.372	860	6.05
Cottonseed oil	–	0.134	4301	2.01
Distiller’s dried solubles	27.4	0.050	1367	6.67
Linseed meal	38	0.551	3077	3.74
Linseed oil	–	0.296	9415	1.76
Molasses (blackstrap)	4.0	0.022	527	0.48
Oat grain	11.8	0.410	1479	2.62
Pea protein	78.8	0.646	4000	45.34
Peas, field	22.0	0.781	1979	7.62
Peanuts (shelled)	25.8	0.753	5260	6.88
Peanut meal	49.1	0.452	3276	7.33
Peanut oil	–	0.301	2184	4.89
Rice grain	8.4	0.218	1673	1.89
Rice hulls	3.0	0.024	335	0.58
Rice, broken kernels	7.0	0.044	187	1.04
Rice bran	13.0	0.022	167	0.52
Rice flour	8.0	0.252	2628	7.16
Rye grain	9.9	0.373	1544	0.313
Sorghum grain	11.1	0.680	3048	5.80
Soybeans	35.2	0.348	2145	3.16
Soybean meal	47.5	0.306	2523	3.46
Soybean protein concentrate	65.0	0.552	3512	35.56
Soybean oil	–	0.077	4190	0.76
Sugar (raw)	–	0.087	1666	1.90
Sunflower seed (dehulled)	23.7	0.684	3366	3.81
Sunflower meal	42.2	0.823	1356	5.81
Sunflower oil	–	0.549	6792	4.20
Vegetable oil (average)	–	0.214	4059	3.65
Wheat grain	11.0	0.295	1829	2.61
Wheat bran	15.7	0.072	462	1.59
Wheat flour	12.1	0.214	1350	4.77
Wheat germ	36.5	0.006	408	0.13
Wheat gluten	79.8	0.036	4189	1.78
Wheat middlings (fine bran)	15.9	0.072	462	1.59
Wheat starch	0.5	0.178	1120	8.85

^aRevision of tables from Chatvijitkul et al. (2017b) and Boyd and McNevin (2022) based on data from Avadi et al. (2015); CPM (1994); Dangol et al. (2015); Gokdogan et al. (2017); Guerin (2020); Lie-Piang et al. (2021); Our World in Data (2020); Ramar et al. (2019); Tripathi et al. (2015); Yani et al. (2022).

Table 3. Embodied land, freshwater, and energy use coefficients for additives commonly included in aquaculture feeds.

Additives	Land (ha/t)	Freshwater (m^3/t)	Energy (GJ/t)
Amino acids (synthetic)	–	100	18.0
Bentonite clay	–	100	2.10
Calcium carbonate	–	100	1.29
Calcium phosphate	–	100	6.96
Casein	0.016	60	0.22
Cellulose	0.220	5732	36.0
Cholesterol	–	100	18.0
Choline	–	100	18.0
Lecithin	–	100	18.0
Mineral premixes	–	100	6.00
Sodium chloride	–	4600	2.80
Sodium phosphate	–	100	8.00
Vitamin C	–	100	18.0
Vitamin premixes	–	100	18.0
Whey (dried)	0.019	547	17.9
Distiller’s dried yeast	0.050	1367	6.67

Source: Boyd and McNevin (2022).

Crude protein concentrations also have been included for feed ingredients (Tables 1 and 2). Crude protein concentrations were from standard tables (FAO 2022; National Research Council 1982; Pork Information Gateway 2010). The sources used are not in complete agreement with respect to crude protein concentrations, because of differences resulting from varieties, agricultural production methods, environmental conditions, and processing techniques. For example, soybean meal from four countries had average crude protein concentrations as follows: Argentina, 45.5%; Brazil, 47.0%; United States, 46.4%, India, 46.3% (Ibáñez et al. 2020). Another example is fishmeal that varies in crude protein concentration depending upon the species from which it was made. Good quality fishmeal has 60–72% crude protein (Cho and Kim 2011). Because of many similar differences, the published tables do not agree exactly on crude protein concentrations for all feed ingredients, but we chose the intermediate concentration or the two concentrations that agreed.

The land, freshwater, and energy use coefficients for plant-based feed ingredients are influenced by the yield per hectare of plant crops. Yields of a crop vary from farm to farm in the same region, among regions, and among countries for numerous reasons such as management, soil fertility, climate and weather. As a simple illustration, in 2021 corn yields ranged from 0.57 t/ha in Zimbabwe to 11.45 t/ha in Turkey with a global average of 5.67 t/ha (USDA, FAS 2023), and in the United States corn production by state ranged from 16.68 t/ha in Washington to 7.06 t/ha in North Dakota with a country average of 11.09 t/ha (USDA 2022). Nevertheless, the average of the highest and lowest estimates, 6.01 t/ha for world production and 11.87 t/ha for United States production are not much different from the averages based on the different countries (5.67 t/ha) or US states (11.09 t/ha). The use of average yields in estimating coefficients is the most practical procedure considering that feed ingredient origin may not be known for any particular batch of feed.

A feed ingredient is typically the result of processing raw material into two or more coproducts, e.g. canola meal and canola oil are coproducts, and fishmeal and fish oil are coproducts. Cottonseed meal is a coproduct of processing seed cotton into cotton lint, seed used for replanting, cottonseed oil, and cottonseed hulls. Therefore, resources used to make feed ingredients must be allocated to coproducts.

Allocation of resource use may be made according to the major use of each crop, e.g. oil for oil crops, lint for cotton, flour for wheat, etc. When done by major use, all the resource use is allotted to the major coproduct for which a crop is considered to be grown and none to the other coproducts. Allocation also can be done based on the economic values of each of the coproducts. For environmental purposes, it seems more logical to allocate the resource uses according to the proportion of the weights of each coproduct to the total weight of all coproducts. Suppose that a yield of 1 ha of a crop was processed and resulted in 0.2 t of A, 0.8 t of B, and 1 t of C (2.0 t coproducts). The land use coefficients would be 0.1 ha/t for A, 0.4 ha/t for B, and 0.5 ha/t for C in accordance with the percentage of each coproduct in the mix. All allocations necessary in calculating the coefficient in Tables 1–3 were made in proportion to the weights of the coproducts.

Embodied land, freshwater, and energy in aquaculture feeds

Earlier calculations of the average amounts of land, freshwater, and energy embodied in aquaculture feed for several common species were based primarily on experimental feed formulations from the literature (Chatvijitkul et al. 2017b; Boyd and McNevin 2022). For the present purposes, generic examples of commercial feed formulae were obtained for use in the present assessment by asking two aquaculture feed consultants who help feed mill operators design their feed formulations to make generic, commercial feed formulae. These included feeds for ictalurid catfish, tilapia, Pangasius, carp, rainbow trout, and whiteleg shrimp. These formulae (Table 4) were used with the updated resource use coefficients (Tables 1–3) to estimate land, freshwater, and energy use per tonne of feed made by each of the generic, commercial feed formulae. Aas et al. (2022) reported the total amounts of feed ingredients used in Atlantic salmon feed in Norway in 2020. These data were used with the updated coefficients from Tables 1–3 to estimate overall values for embodied land, freshwater, and energy embodied per tonne of Atlantic salmon feed.

Table 4. Generic, commercial feed ingredient formulae with percentage inclusion rates of each ingredient.^a

Ingredient	Tilapia^a	Ictalurid Catfish^a	Pangasius^a	Carp^b	Rainbow Trout^a	Whiteleg shrimp^a
Blood meal	–	5.0	–	–	4.0	–
Fishmeal, marine	–	–	–	–	15.0	6.0
Fishmeal, rendered	–	–	–	–	–	–
Fish oil, marine	–	–	–	–	8.5	3.2
Fish oil, rendered	1.8	3.1	3.0	1.3	–	–
Fish solubles	–	–	–	–	–	2.5
Meat and bone meal	7.0	7.0	–	–	–	–
Poultry byproduct meal	–	–	–	–	7.0	–
Shrimp head meal	–	–	–	–	–	2.4
Cassava flour	–	10.0	14.4	–	–	–
Corn gluten meal	–	2.0	–	8.9	6.0	–
Corn grain	18.0	–	–	–	–	–
Rice bran	–	16.9	14.0	16.0	–	–
Rice, broken	–	–	11.0	–	–	–
Soybean meal	55.0	36.0	55.0	50.0	–	52.6
Soybean oil	–	–	–	–	3.0	–
Soybean protein concentrate	–	–	–	–	30.0	–
Wheat flour	–	17.0	–	–	16.0	28.2
Wheat grain	15.0	–	–	21.0	–	–
Wheat middlings	–	–	–	–	7.3	–
Antioxidant	0.05	0.01	–	–	0.05	0.03
Carophyll pink	–	–	–	–	0.04	–
Choline chloride	0.01	–	–	–	0.03	0.06
DL-methionine	0.14	0.12	0.13	–	0.1	–
Ethoxyguin	–	–	–	0.02	–	–
Hydrolyzed fish attractant	–	–	–	–	1.0	–
Mold inhibitor	0.05	0.05	0.05	–	0.5	0.03
Monocalcium phosphate	1.20	0.95	1.51	2.43	–	1.20
Mycotoxin binder	0.10	0.10	0.10	–	0.10	0.20
Pellet binder	–	–	–	–	–	0.4
Premix, mineral	0.25	0.25	0.25	.25	0.25	0.25
Premix, vitamin	0.50	0.50	0.50	0.1	0.5	0.40
Soybean lecithin	1.0	–	–	–	1.0	1.5
Stay-C	0.03	–	–	–	–	0.04
Vitamin C	–	0.03	0.03	–	0.03	–-

^aCourtesy Mark Newman, Aquaculture Feed Consultant.

^bCourtesy Allen Davis, Auburn University.

The ingredients for feeds must be transported to feed mills by energy consuming conveyances, feed mills use energy to mix the ingredients and produce the pelleted feed, and energy is required to deliver the feed. Although energy use for feed is a component of several aquaculture LCAs, only one LCA report was found in which energy use associated with the feed mill phase could be separated with certainty from other energy uses. Avadi et al. (2015) reported that feed mill-associated energy use was 1.12 GJ/t for rainbow trout feed and 0.682 GJ/t for tilapia and black pacu feeds. The estimate for catfish feed in the United States was 1.07 GJ/t (Boyd et al. 2008). The average of these values, 0.89 GJ/t used here is not greatly different from the value of 1.09 GJ/t used by Boyd and McNevin (2022) in calculating the embodied energy in aquaculture feeds.

Land use for feed mill operations was assumed to be similar to that of fishmeal plants and animal byproduct-rendering plants (see Table 1). Freshwater use at feed mills was taken as 50 m³/t feed.

The generic commercial feed formulations for all species but carp required more land and all of them required more freshwater than reported earlier by Boyd and McNevin (2022) for feeds for the same species based mainly on experimental feed formulae from the literature. Embodied energy was lower for all species other than Atlantic salmon in the commercial feeds than in the experimental feeds (Table 5). The reasons for these differences will not be discussed because from a practical standpoint, very little of the feed used in aquaculture results from experimental feed. The reasons are various to include differences in ingredients used, the target crude protein contents, and inclusion rates of fishmeal, animal byproduct meals, and plant protein substitutes for animal meal. Another important reason is that the revised feed ingredient resource use coefficients (Tables 1 and 2) were applied to the commercial feed formulas. There were some large reductions in energy use coefficients of fishmeal, fish oil, and other animal source feed ingredients in the revised tables as a result of information not previously available.

Table 5. Land, freshwater, and energy embodied in 1 tonne of commercial feed for seven common aquaculture species compared with average (n = 7), embodied resource use estimates made earlier for experimental diets for six of the species.

Feed^a	Land (ha/t)	Water (m^3/t)	Energy GJ/t)	Crude protein (%)
Tilapia
Commercial	0.273	2061	4.50	32.1
Experimental	0.200	1405	5.47	30.0
Ictalurid catfish
Commercial	0.233	2231	4.11	32.0
Experimental	0.201	1574	4.78	29.8
Pangasius
Commercial	0.259	1793	4.35	28.1
Experimental	0.227	1553	5.13	29.5
Carp
Commercial	0.220	1803	3.54	33.8
Experimental	0.226	1552	6.30	30.0
Atlantic salmon
Commercial	0.292	2842	14.19	37.0
Experimental	0.173	1780	13.65	39.9
Rainbow trout, commercial	0.232	1771	14.54	45.1
Whiteleg shrimp
Commercial	0.221	1726	5.79	35.1
Experimental	0.202	1612	9.59	34.2

^aFor commercial diets resource use amounts calculated from feed formulae in Table 4; and data on Atlantic salmon feed ingredients from Aas et al. (2022) the resource use amount for experimental diets taken from Boyd and McNevin (2022).

Among the generic, commercial feeds (Table 5), land use ranged from 0.220 ha/t for carp to 0.292 ha/t for Atlantic salmon. The high embodied land content of Atlantic salmon feed as compared with carp feed resulted because most of the crude protein included in carp feed was derived from soybean meal and corn gluten meal which have comparatively low embodied land contents as compared to the large amount of land embodied in soybean protein concentrate (Table 2) that made up 21% of feed ingredients in the Atlantic salmon feed. The commercial feed for tilapia also had a relatively high land use of 0.273 ha/t, and 61.5% of the land use was for soybean meal.

Embodied freshwater contents of feeds were calculated using the total embodied freshwater footprints of feed ingredients (Table 2) that include green water (rainfall) which evaporates or is transpired from lands under cropping (Mekonnen and Hoekstra 2011). Total water footprints for agricultural products usually are about five-fold greater than consumptive freshwater use that consists only of freshwater withdrawn from underground sources and streams (Boyd and McNevin 2015). From an environmental viewpoint, the total freshwater footprint is thought to be the best indicator of human interventions into the hydrologic cycle. In coastal and marine aquaculture, ocean or estuarine water used directly in production is not part of the renewable, available freshwater component of the hydrological cycle used for human purposes (Gleick 2019). In marine and brackishwater aquaculture, freshwater use results mainly from embodied freshwater in feed (Boyd et al. 2021a).

Freshwater use for the generic, commercial feed formulations ranged from 1726 m³/t for whiteleg shrimp feed to 2842 m³/t for Atlantic salmon feed (Table 5). Atlantic salmon feed had 661 m³/t more embodied freshwater than the next lowest feed for ictalurid catfish. The large amount of freshwater use for salmon feed resulted primarily from the substitution of soy protein concentrate and rapeseed oil, ingredients of particularly high embodied freshwater coefficients, as replacements for fishmeal and fish oil. Feed mill operators tend to be lowering fishmeal and fish oil inclusion rates through the use of animal byproduct meal, oil seed meals, vegetable oils, and soybean and pea protein concentrates (Aas et al. 2022; Naylor et al. 2021).

Embodied energy use, 14.19 GJ/t and 14.54 GJ/t, in the commercial formulations for Atlantic salmon and rainbow trout feed, respectively, was greater than in the other feeds (Table 5). The two salmonid feeds had high inclusion rates of fishmeal, fish oil, and plant protein concentrates of markedly higher embodied energy coefficients than incurred by the other feeds. The embodied energy contents of fish feeds and whiteleg shrimp feed were between 3.54 GJ/t and 5.79 GJ/t.

Wild fish use in feeds

Aquaculture feed production consumes most of the annual production of fishmeal and fish oil. There is much concern over the sustainability of world fisheries, and it is common to calculate the efficiency with which wild-caught fish used for feed ingredients are converted to aquaculture biomass or the fish in–fish out ratio or FIFO (Boyd and McNevin 2015). The FIFO ratios were calculated initially in the present assessment by an equation developed by Boyd and McNevin (2015). Fishmeal and fish oil rendered from fish and other aquatic animal processing waste included in feed are not considered in the FIFO calculation that is based only on fishmeal and fish oil made from whole, small pelagic fish. As a result, the FIFO was 0.0 for tilapia, ictalurid catfish, carp, and Pangasius feeds. The FIFO values in tonnes of wild fish used per tonne of aquaculture biomass harvested for the other species were: whiteleg shrimp, 1.16; Atlantic salmon, 2.45; rainbow trout, 2.51.

There are other ways of calculating the FIFO. The Aquaculture Stewardship Council (ASC) calculates it by another name, the forage fish dependency ratio (FFDR). The method involves calculating the FFDR for fishmeal and fish oil separately and using the larger of the two values as the FFDR for a particular situation (ASC 2019). The equations are: (1) $$\mathrm{\text{FFD}}{\mathrm{R}}_{\mathrm{m}}=\frac{\mathrm{\text{FM}}\times \mathrm{\text{FCR}}}{24}$$(1) and (2) $$\mathrm{\text{FFD}}{\mathrm{R}}_{\mathrm{o}}=\frac{\mathrm{\text{FO}}\times \mathrm{\text{FCR}}}{5,6,\text{or}7}$$(2) where: FFDR_m = forage fish dependency ratio based on fishmeal, FM = marine fishmeal included in feed (%), FO = marine fish oil included in feed (%), FFD_o = forage fish dependency based on fish oil, FCR = quantity of feed used to produce the amount of fish harvested, 24 = 24% fish meal from live forage fish, 5 = 5% fish oil from live forage fish from Peru, Chile, and Gulf of Mexico fisheries, 7 = 7% fish oil from forage fish from Denmark, Norway, Iceland, and UK fisheries. The 6 = 6.0 fish oil in instances where the source of the fish is unknown (this is implied but not stated in the ACS instructions).

The Global Aquaculture Alliance (GAA) Best Practices Certification (BAP) farm standard (BAP 2022) uses a different methodology: (3) $$\mathrm{\text{FFIF}}=\frac{\mathrm{\text{FM}}+\mathrm{\text{FO}}}{22.5+4.8}$$(3) (4) $$\mathrm{\text{FIFO}}=\mathrm{\text{FFIF}}\times \mathrm{\text{FCR}}$$(4) where: FFIF = feed fish inclusion factor, 22.5 = 22.5% fishmeal from live forage fish, 4.8 = 4.8% fish oil from live forage fish (irrespective of fishery).

The International Feed Fish Organization (FIFO) gives yet another FIFO method (Jackson 2009), it follows: (5) $$\mathrm{\text{FIFO}}=\frac{\mathrm{\text{FM}}+\mathrm{\text{FO}}}{22.5+5.0}\times \mathrm{\text{FCR}}$$(5)

The symbols have been defined above, and 22.5 and 5.0 are the percentage recoveries of meal and oil from forage (or feed fish), respectively. This equation is very similar and likely the source of the BAP equations above.

The FIFO in the ASC depends on whether the most live fish is needed for fishmeal or for fish oil. The weighted average of fishmeal and fish oil are used to estimate live fish in the BAP and IFFO methods. There would seem to be an issue related to the excess of either fishmeal or fish oil derived from the amount of live fish accounted to the quantity of wild fish calculated as used in aquacultural production because the excess meal or oil can be used for another purpose.

The equation developed by Boyd and McNevin (2015) attempted to adjust to some extent for this issue. The original equation in which the percentage recoveries of 20% meal and 5% oil were adjusted to those recoveries used in the ASC equation. These adjustments can be justified in order to comply with recovery rates reported in more recent sources (Boyd et al. 2022) than were used in 2015. At the new recovery rates, 1 kg fishmeal requires 4.17 kg live fish, a ratio of 0.239 of meal to fish, but results in only 0.25 kg of fish oil, a meal to oil ratio of 4.0. Thus, if the fishmeal content of the feed is 4.0 or greater, no additional live fish would be required for the oil. The adjusted equation is: (6) $$\mathrm{\text{FIFO}}=\left[\frac{\mathrm{\text{FM}}}{\left(100\right)\left(0.239\right)}\right]\times \mathrm{\text{FCR}}$$(6) where FM = inclusion rate equals or exceeds 4.0.

If the fishmeal to fish oil inclusion rate is below 4.0, the equation to use is: (7) $$\mathrm{\text{FIFO}}=\left[\left(\frac{\left(\frac{\mathrm{\text{FO}}}{100}\right)‐\mathrm{ }\left(\frac{\mathrm{\text{FM}}}{100\times 4.0}\right)}{0.06}\right)+\left(\frac{\mathrm{\text{FM}}\times 4.17}{100}\right)\right]\times \mathrm{\text{FCR}}$$(7) where 100 = the factor to convert percentages to decimal fractions,0.06 = ratio of fish oil to live fish,0.239 = ratio of fish meal to live fish,4.17 = ratio of live fish to fishmeal.

The results of the calculations made using the four equations are shown in Table 6. In making these calculations, the data from Aas et al. (2022) for Norwegian feed ingredients used in 2020 for salmon feed in Norway were used. For these reported amounts of feed ingredients versus feed produced, the fishmeal and fish oil from whole marine fish were 8.81% and 8.33%, respectively. The amounts for rainbow trout and shrimp were taken from Table 4.

Table 6. Results for the fish in—fish out ratio (FIFO) by different methods of calculation.

Method	Whiteleg shrimp	Rainbow trout	Atlantic salmon
ASC	0.85	1.84	1.80
BAP	0.54	1.12	0.82
IFFO	0.54	1.11	0.81
Boyd and McNevin^a	1.16	2.45	2.51
Boyd and McNevin, adjusted^b	0.85	1.85	1.81
Boyd and McNevin, adjusted^c	0.79	1.79	2.25
Kok et al.^d	0.52	0.98	0.98

^aOriginal equation assumed 20% recovery of fishmeal and 5% recovery of fish oil from live fish (Boyd and McNevin 2015).

^bAdjusted for recoveries of 24% fishmeal and 6% fish oil from live fish as per ASC.

^cAdjusted for recoveries of 22.5% fishmeal and 4.8% fish oil from live fish as per BAP and similar to IFFO.

^dKok et al. (2020). The data from this study is for the economic FOFO and based on the global production and feed use by the species considered. The eFiFO is not for a specific feed formulation.

The BAP and IFFO methods differ only in that the percentages recovery of fish oil from wild fish was slightly different between them. As a result, they give essentially identical FIFO ratios in all comparisons (Table 6). The ASC method resulted in markedly greater FIFOs than those calculated by the BAP and IFFO equations for the three species. The original Boyd and McNevin equation gave much greater FIFOs than did the BAP and IFFO equations, and appreciably higher FIFOs than did the ASC method. The adjusted Boyd and McNevin equation to achieve better agreement with recent estimates of the percentage of fishmeal from live fish, and the adjustment of fish oil recovery based on the fishery source of the fish for making just oil resulted in FIFOs very similar to the ASC results. To be consistent, the Boyd and McNevin equation was adjusted to recoveries of 22.5% fishmeal and 4.8% fish oil from live fish as done by BAP and which is only 0.2% less than the recovery of 5% fish oil assumed by the IFFO method. This adjustment gave considerably greater FIFO ratios than obtained by the BAP and IFFO equations.

The conclusions are several. The FIFO is sensitive to the fishmeal and fish oil recovery rates used in the equations, which resulted from the actual recovery rates for a particular species of live marine fish used in making the meal and oil. Moreover, the source of the fishmeal may not always be known. The Boyd and McNevin equation and the method used by ASC are both superior and more reasonable as a certification requirement because they make a more accurate estimate of the actual amount of wild fish used to give the desired inclusion rates for both ingredients. The other two methods made an average calculation that is influenced greatly by the ratios of the two ingredients. Finally, the FCR is an extremely important factor as it determines the amount of feed needed per unit of biomass yield. Thus, it greatly influences the FIFO.

Kok et al. (2020) used an economic allocation to quantify the FIFO for major fed aquaculture species. The method they used resolves economically the issue of there being either excess fishmeal or fish oil use for other purposes in the FIFO calculation. The calculation requires considerable economic data, and the equation for calculating the economic FIFO can be found in Kok et al. (2020). The results that were reported for marine shrimp, Atlantic salmon, and rainbow trout are given in Table 6. They also reported eFIFO values of 0.06 for tilapia, 0.09 for catfish (mainly Pangasius presumably), and 0.05 for fed carp (Table 6). The eFIFO values are in most cases lower than the ones calculated by other methods.

It was stated earlier that the authors felt that resource use coefficients should not be based on economic allocations but on actual amounts of resources used. The use of the eFIFO appears to be another place where there may be disagreement over the environmental aspect and the economic aspect of resource use.

Resource sparing through feed ingredient selection

The amounts of embodied land, water, and energy were similar for the generic, commercial feed among the four, non-salmonid fish and shrimp (Table 5). The averages and standard deviations were as follows: land, 0.247 ± 0.028 ha/t; freshwater, 2032 ± 402 m³/t; energy, 4.46 ± 0.83 GJ/t. The non-salmonid fish feeds did not require fishmeal or fish oil produced from whole, marine fish, but shrimp feed required 540–1160 kg of whole, marine fish/t, depending on which equation was used in the calculation (Table 6). The salmonid feeds were slightly higher than the fish and shrimp feeds with respect to embodied land and water use averaging 0.262 ha/t and 2306 m³/t, respectively. Embodied energy use was much greater for salmonid feeds (average = 14.80 GJ/t) than for the non-salmonid fish and shrimp feeds. Of course, there was much use of whole, marine fish in salmonid feeds and little difference between rainbow trout and Atlantic salmon.

The amounts of embodied land, freshwater, energy, and wild fish in feeds could be controlled to some extent at feed mills by the choice of ingredients. Two examples are given for the amounts of embodied resources and quantities of different ingredients necessary for 10 kg (1% inclusion rate) of crude protein in 1 t of feed (Tables 7 and 8). The calculations were made as follows: $$\mathrm{E}{\mathrm{R}}_{\mathrm{i}}=\frac{\mathrm{C}{\mathrm{P}}_{\mathrm{a}}}{\mathrm{C}{\mathrm{P}}_{\mathrm{\text{ci}}}}\times {\mathrm{k}}_{\mathrm{i}}\left(8\right)$$ where ER = embodied resource used (ha land/t; m³ water/t; GJ energy/t), CP_a = amount of crude protein included (t), CP_ci = decimal fraction of CP concentration in ingredient (% ÷ 100), k_i = appropriate embodied resource coefficient.

Table 7. Embodied land, freshwater, and energy use resulting from inclusions of selected grain products to provide to provide 1% (10 kg) of crude protein in 1 t of aquaculture feed.

Ingredient	Quantity (t/t)	Land (ha/t)	Freshwater (m^3/t)	Energy (GJ/t)
Corn meal	0.111	0.028	120	0.53
Rice flour	0.125	0.032	328	0.90
Sorghum grain	0.090	0.061	274	0.28
Wheat bran	0.064	0.005	30	0.10
Wheat flour	0.083	0.018	112	0.40
Wheat middlings	0.063	0.004	29	0.10

Table 8. Embodied land, freshwater, and energy use resulting from inclusions of selected oil plant meals to provide 1% (10 kg) of crude protein in 1 t of aquaculture feed.

Ingredient	Amount (t/t)	Land (ha/t)	Freshwater (m^3/t)	Energy (GJ/t)
Canola (rapeseed) meal	0.0281	0.008	53	0.174
Coconut meal	0.0476	0.035	100	0.132
Linseed (flax) meal	0.0263	0.014	81	0.098
Peanut meal	0.0204	0.008	18	0.123
Soybean meal	0.0211	0.006	53	0.073

The amounts of embodied land, freshwater, and energy associated with 1% crude protein in aquaculture feeds were calculated. Wheat middlings and wheat bran resulted in considerably less of all three embodied resources than did the other grain products (Table 7). Cornmeal and rice flour resulted in nearly twice as much land as did wheat flour. Wheat bran and wheat middlings required roughly one-fourth as much embodied land as did wheat flour. The embodied resource use resulting from the grain product ingredients can be summarized as follows:

• Land: wheat middling ≈ wheat bran < wheat flour < cornmeal ≈ rice flour < sorghum grain;

• Water: wheat middling ≈ wheat bran < wheat flour ≈ cornmeal < sorghum grain < rice flour;

• Energy: wheat bran = wheat middling < sorghum grain < wheat flour < cornmeal < rice flour.

Soybean meal and other oil meals (Table 8) required about two- to four-fold more land than did the grain products. Some of the oil meals were lower with respect to freshwater and energy. Embodied resource use was in the following order for the oil meals:

• Land; peanut ≈ soybean = canola < linseed < coconut

• Water; peanut < soybean = canola < linseed < coconut

• Energy; soybean < linseed < peanut ≈ coconut < canola.

Coconut meal incurred the greatest amounts of embodied land and freshwater, while canola meal resulted in the most embodied energy. Soybean meal seems to be the lowest in embodied resource use overall, and this is fortunate because this meal is widely used in aquaculture feeds.

The estimates (Tables 7 and 8) did not consider ingredient cost or protein quality. Feed mill operators usually select feed ingredients based on availability, compatibility for compounding feeds, protein quality, and with particular attention to cost. The emphasis is on least cost formulations that have both acceptable nutritional quality and pellet stability. At present, the only effort at lessening embodied resource use in aquaculture feeds seems to be directed at replacing all or a part of their whole fish-derived fishmeal and fish oil contents with rendered fishmeal and fish oil or high crude protein content plant products.

The amounts of embodied resource use incurred per unit weight of fish or shrimp produced in feed-based aquaculture are directly proportional to the embodied resource use coefficients per tonne of feed and the quantity of feed required per unit of biomass harvested. The FCR is an important farm-level factor related to resource use conservation to include the potential negative impacts associated with feed use. As already mentioned, lowering FCR by only 0.1 would lessen feed use per tonne biomass harvested by 100 kg.

Suppose a feed manufacturer had been producing a feed that contained 100 kg/t of sorghum grain and replaced it with 100 kg of wheat grain which is equal to sorghum in crude protein concentration. The embodied resource content of feed with wheat would be less in embodied resources by 0.0385 ha land/t, 122 m³ freshwater/t, and 0.319 GJ energy/t. Assuming the wheat protein is about the same quality as the sorghum protein, the wheat-containing feed should allow the same FCR as possible with the feed containing the sorghum.

Suppose that the farm using the wheat-containing feed adopts better feeding practices and obtains an FCR that is 0.1 unit lower. This would reduce feed input per tonne of production by 100 kg. The reduction in the embodied resource use achieved that would be associated with aquaculture production at the farm level would be equal to that achieved by replacing 100 kg sorghum with 100 kg wheat at the feed mill. This example was designed for simplicity of illustration, but the implication is that the reduction in embodied resource use at the farm level can be as great or greater than that possible by selecting feed ingredients based on their embodied resource coefficients at the feed mill.

Feed mill operators have a goal to manufacture feed with an adequate nutrient content for the target species according to the least cost of ingredients. This practice is called in the trade least cost formulation. Several factors are involved in choosing ingredients to include nutrient content, digestibility and especially the protein availability in ingredients, and cost of each. When substitutions of marine fishmeal and marine fish oil are made, there may or may not be a reduction in cost. The reduction in marine fishmeal and fish oil through substitutions of other feed ingredients would not likely be done by a feed mill operator if the cost of the finished feed would be more expensive. Of course, if there is a market demand for feed with lower marine fishmeal and fish oil inclusion rates, even at a higher price, then feed mill operators would most likely be willing to produce that feed.

Putting aside the economic aspect of feed ingredient selection, the possibility of sparing resources by choice of feed ingredients will be illustrated. Without doing a lot of calculations to adjust crude protein concentration exactly, the embodied resource use coefficients in Tables 1 and 2 were used to determine the changes in embodied land, water, and energy that would result in a shrimp feed containing 8% marine fishmeal and 3% marine fish oil when 50% of the marine products were replaced with soy protein concentrate and canola oil or with pig byproduct meal and animal oil.

The original feed formula had the following amounts of embodied resources that resulted from the two marine products: land, 0.00005 ha/t; water, 8 m³/t; energy, 0.905 GJ/t. The feeds modified by substitution had the following embodied resources associated with the remaining fishmeal and fish oil, plus the substituted products: plant product substitution (0.025 ha/t land, 209 m³/t water, and 1.97 GJ/t energy); animal byproduct substitution (0.012 ha/t land, 162 m³/t water, and 0.625 GJ/t energy).

The substitutions resulted in increases in land, water, and energy use, but of course, the amounts of fishmeal and fish oil were cut in half. Resource use by the plant product substitutions were calculated to be about double for land, one-fourth greater for water, and three-fold larger for energy than with the animal byproduct substitutions. The amount of the increases in embodied resources in harvested biomass is in direct proportion to the FCR achieved. The FIFO would decrease by 50% at whatever FCR was achieved.

The potential for reducing resource use and negative environment impacts directly associated with feed cannot be realized fully without efforts at both feed mills and farms. Catfish farming in the United States is a good example of an unnecessarily high FCR. The farmers employ a feeding practice of applying a certain amount of feed by mechanical feed blowers with little regard to the amount eaten (this is based on the senior author’s personal observations over five decades). The fish are produced in a multiple-harvest system in which ponds are not drained for harvest and fish are harvested with large tractor-drawn seines. Some fish evade capture and grow larger and less efficient in converting food to biomass (Boyd and Tucker 2014). The average FCR is around 2.5 as estimated from the total production and total feed use reported to USDA annually and available in issues of the Catfish Journal (www.catfishfarmersofamerica.com). Studies conducted over many years on research stations in which ponds were drained annually for harvest seldom have recorded FCRs above 2.0, and most were around 1.4–1.8 (Boyd and Tucker 2014). In the case of this species, there is a great opportunity to lessen embodied resource use and diminish feeding waste through improvements in pond management techniques that can lessen FCR. Of course, there is opportunity to reduce FCR across all the different types of feed-based aquaculture.

Energy use by aeration

Aerators are used in many kinds of feed-based aquaculture, and their oxygen-transfer rates usually are in the range of 0.8–2.0 kg O₂/kW·hr (Boyd and McNevin 2021). Oxygen movement from air to water decreases as dissolved oxygen concentration in water increases, but aerator test results are reported as the capabilities of aerators to transfer oxygen from air to completely oxygen-depleted water. Dissolved oxygen concentrations usually are above 3–4 mg/L in culture of warmwater and tropical species and above 5–6 mg/L for coldwater species (Boyd and Tucker 1998). These concentrations represent about 40–60% of dissolved oxygen saturation, and aerators will usually not transfer more than about 50% of their standard aeration transfer rates that were measured at 0.0 mg/L dissolved oxygen, 20 °C, and clean water in aerator performance tests (Boyd and Tucker 1998).

Using whiteleg shrimp as an example, the production of 1 t of biomass would impose a total BOD from feed oxidation of 1646 kg O₂. A crop duration of 80 d is typical, and the aerators in shrimp ponds usually are operated about 18 hr/day or 1620 hr/crop (Boyd et al. 2021a). Electric aerators normally used by shrimp farmers have small motors (1–3 hp) that use electricity at an efficiency of around 75% (Hargreaves and Boyd 2022). One horsepower is roughly 75% of 1 kW·hr (1 kW = 1.341 hp or 1 hp = 0.746 kW), and each hour of aerator operation equates almost exactly to 1 kW·hr of electricity use (1 hp·hr = 0.746 kW·hr ÷ 0.75 motor efficiency = 0.995 kW·hr). Small aerators have a standard oxygen transfer rate of about 1 kg O₂/hp (Boyd and McNevin 2021), and they will probably transfer about half this much oxygen under actual operating conditions. This suggests that 1 hp of aeration in a shrimp pond will supply 810 kg O₂/crop. The total BOD for feeding white leg shrimp was estimated to be 1646 kg O₂/t production or 2.03 hp/t. As a result, the shrimp biomass to installed aerator capacity ratio would be 0.49 t/hp of shrimp per crop.

Shrimp farmers in Asia typically assume that 1 hp of aeration by small floating, electric paddlewheel aerators will support about 400–500 kg shrimp. The calculation of the ratio from the total BOD done here merely verifies what shrimp farmers have learned through experience. Of course, long-arm paddlewheel aerators fabricated onsite by shrimp farmers in Asia are not as efficient as factory-made long-arm aerators or normal floating, electric paddlewheel aerators. In Thailand, farmers usually apply a production to installed aerator capacity ratio of 250–350 kg shrimp/hp when using farm-made long arm aerators (Boyd and McNevin 2021).

The amount of aeration also can be put on a feed basis. At a production to aerator capacity ratio of 400–500 kg shrimp/hp, feed input at a typical feeding rate of 2% body weight per day would be 8–10 kg feed/ha/d. Feeding rates per unit of body weight present are higher early in the shrimp crop and less near the time of harvest. Boyd and McNevin (2021) suggested that shrimp farmers should monitor dissolved oxygen concentrations in ponds and adjust the amount and time of aeration according to actual need. To illustrate, in a shrimp pond with production of 8 t/ha/crop, 20 hp of aeration might be installed. Over an 80-day crop, a reduction in aerator run time from an average of 16 hr/d to 14 hr/d would lessen electricity use by 3200 kW·hr or 1.44 GJ/t shrimp.

Energy use per tonne of production for aeration in fish culture should be similar to that of shrimp as oxygen consumption rates in respiration are similar (Boyd and Tucker 1998). In the shrimp culture example above, 490 kg shrimp/1620 kW·hr per crop equates to 3306 kW·hr/t shrimp or 11.9 GJ/t shrimp. This is direct energy use; the efficiency with which electricity is generated from primary energy varies among electrical grids but on average is around 40% (Table 9). Thus, 11.9 GJ/t direct energy would result in 29.8 GJ/t total energy. The embodied energy in shrimp feed at an FCR of 1.6 would be only 9.3 GJ/t shrimp harvested.

Table 9. Direct energy content and embodied resource content of common energy sources used in aquaculture.

		Direct energy	Embodied resource content
	Unit	content (GJ)	Land (ha)	Freshwater (m^3)	Energy (GJ)
Electricity^a	kW·hr	0.0036	1.6 × 10^−6	0.008	0.00539
Diesel fuel^b	L	0.0387	NF	0.0003	0.00701
Gasoline^c	L	0.0349	NF	0.0004	0.0084
Liquid propane (LPG)^c	L	0.0261	NF	NF	0.0003

^aShapouri et al. (2002; Stevens (2017); Torcellini et al. (2003).

^bSun et al. (2018).

^chttp://www.iea.org/statistics/resources/manuals/. . . .

Catfish farming in the United States stands out as particularly inefficient with respect to feed use. With respect to mechanical aeration, the large, 10-hp paddlewheel aerators used in catfish ponds have standard aeration rates of 1.8 kg O₂/hp·hr as compared to around 1.0 kg O₂/hp·ha for the small, 1-hp to 3-hp floating, electric and by long-arm aerators powered by 1- to 3-hp electric motors or 8- to 14-hp diesel engines mounted on pond dikes used in shrimp farming in Ecuador and in several Asian countries (Boyd et al. 2018; Boyd and McNevin 2021; Hargreaves and Boyd 2022). Aeration is applied in catfish ponds for about 5 months during the year and usually around 10 hr each day (at night). This would result in 1500 hr aeration annually. By the same reasoning applied to shrimp pond aeration above, direct aeration energy use in US catfish farming is about 10.0 GJ/t and total energy use is 24.9 GJ/t. With respect to energy use, catfish farmers use less energy to produce 1 t biomass than do shrimp farmers who achieve a lower FCR. The implication is that improvements can be made in both types of culture: in one by improving feed conversion, in the other by using more efficient aerators.

It is important to consider the embodied energy in fuels, and particularly when comparing farms using diesel fuel to generate end-use energy onsite with those using grid electricity that is made from primary fuels at an offsite generating plant. The average embodied land, freshwater, and energy contents of major energy sources used in aquaculture are provided in Table 9. For electricity, 1 kW·hr equates to 0.0036 GJ; thus, the embodied energy content results in an additional input of 0.00539 GJ of primary energy for generation and delivery through the grid. In other words, 1.497 times more energy than delivered is used in providing end use electricity. Diesel engines also may be used at aquaculture farms, but they are considerably less efficient in converting input energy in fuel to end use energy than are electric motors in transforming input electricity to useful energy. The embodied energy content of diesel fuel is less than for electricity, and only adds about 0.18 times more energy than is contained in the fuel.

Typical 2–3 hp motors used to power shrimp pond aerators require about 1 kW·hr to provide 1 hp·hr of energy at the aerator shaft (Hargreaves and Boyd 2022). A small, naturally aspirated, diesel engine requires about 0.21 L fuel to provide 1 hp·hr (Barrington Diesel Club, Undated). One horsepower·hour of end-use energy from electricity requires a total of 0.00899 GJ total energy, while the diesel engine can do the same using 0.0097 GJ total energy.

This suggests that diesel is only slightly less energy efficient than electricity. However, diesel engines require more maintenance and servicing than do electric motors, and at recent world average prices of electricity of US $0.165/kW·hr and diesel fuel of US $1.28/L (Anonymous 2022, 2023), 1 hp·hr of end-use energy would cost US $0.165 by electricity and US $0.218 by diesel fuel. Where grid electricity is available, it is usually preferred over diesel fuel by fish and shrimp farmers. Of course, diesel generators can be used onsite to produce electricity. A 1000-kW generator yields about 3.8 kW/L diesel, an energy efficiency of 35.35%. To provide 1 kW·hr necessary for a 1 hp-hr energy input to an aerator shaft would require 0.763 L fuel or a total energy use of 0.0117 GJ (Generator Source, Undated). This is 0.002 GJ more than required by diesel aerators and 0.0027 GJ more than necessary to electric aerators for each 1-hp·hr of aerator operation.

System waste loads

The waste loads of carbon, nitrogen, and phosphorus released into the water of the culture system directly as a result of feed are the differences in the amounts of each of these three key elements added in the feed minus those portions of the elements removed in harvested biomass. The equation for calculation of the elemental waste loads per tonne of biomass is: (9) $$\mathrm{W}{\mathrm{L}}_{\mathrm{i}}=\left(\mathrm{F}\times {\mathrm{E}}_{\mathrm{i},\mathrm{f}}\right)-\left(1000\times {\mathrm{E}}_{\mathrm{i},\mathrm{b}}\right)$$(9) where WL_i = waste load of element_i (kg/t harvest biomass), F = feed input (1000 × FCR), E_i,f = decimal fraction of element_i in feed (% ÷ 100), 1000 = kg/t, E_i,b = decimal fraction of element_i in harvest biomass (% ÷ 100).

The values of carbon, nitrogen, and phosphorus concentrations in feed and biomass, the FCRs used in the calculations, and the estimated system waste loads are found in Table 10. The whole body biomass of the culture species had similar concentrations of each of the three elements. FCR values ranged from 1.3 for Atlantic salmon and rainbow trout to 2.5 for ictalurid catfish. The FCRs for the other fish species and whiteleg shrimp were between 1.5 and 1.7. The concentrations of carbon in feed ranged from 38.58% for carp to 46.46% for salmon. Nitrogen concentration ranged from 5.12% for ictalurid catfish to 7.22% for rainbow trout. Carbon concentration increased with respect to greater amounts of oil and crude protein in feed, and the nitrogen concentration was in direct proportion to crude protein concentration. Phosphorus concentration varied from 1.09% in carp and whiteleg shrimp feeds to 1.45% in Pangasius feed.

Table 10. System waste loads of carbon C_L, nitrogen (N_L), and phosphorus (P_L) from feed-based culture of 1-t harvest biomass of seven common aquaculture species at typical feed conversion ratios (FCR).

		Feed (%)^b			Biomass (%)^c			System waste load (kg/t biomass)
	FCR^a	C	N	P	C	N	P	CL	NL	PL
Carp	1.7	38.58	5.41	1.09	10.5	2.42	0.38	551	67.8	14.7
Ictalurid catfish	2.5	40.97	5.12	1.21	10.5	2.38	0.68	919	104.2	27.1
Tilapia	1.7	39.16	5.14	1.19	10.9	2.22	0.70	557	65.2	13.7
Pangasius	1.5	36.37	4.50	1.45	11.0	2.34	0.59	436	44.1	15.8
Atlantic salmon	1.3	46.46	5.92	1.43	12.0	2.96	0.32	484	47.4	15.4
Rainbow trout	1.3	43.24	7.22	1.39	12.0	2.50	0.35	442	68.9	14.6
Whiteleg shrimp	1.6	39.28	5.62	1.09	11.7	2.86	0.40	512	61.3	10.2

^aIctalurid catfish estimated from US production and feed use in 2021 (Hanson 2022); others from Naylor et al. (2021).

^bChatvijitkul et al. (2018).

^cCarp (Sterne and George 2000); others (Boyd et al. 2007).

The system waste loads of carbon per tonne of harvest biomass for fed fish and shrimp ranged from 436 kg/t of Pangasius to 919 kg/t of ictalurid catfish (Table 10). Those for nitrogen varied from 44.1 kg/t of Pangasius to 104.2 kg/t of ictalurid catfish, while for phosphorus the lowest was 10.2 kg/t of whiteleg shrimp and the greatest phosphorus load was 27.1 kg/t of ictalurid catfish. The ictalurid catfish system waste loads are in a separate category and this attests to the high FCR realized by US catfish farmers as compared to the producers of other species as already alluded to above. The ictalurid catfish grown in the United States can be produced with FCRs of 1.6–2.0 as revealed by numerous studies, the results of which have been summarized by Boyd (1990) and Boyd and Tucker (1998, 2014).

Ictalurid catfish, excluded, the ranges in system waste loads were 442–557 kg/t of carbon, 44.1–68.9 kg/t of nitrogen, and 10.2–15.8 kg/t of phosphorus (Table 10). It should be noted that FCR exerts considerable influence on system waste loads. Some reduction in system waste loads of nitrogen and phosphorus could be realized by including as little of these two elements in feeds as nutritional allowable without affecting growth (Gross et al. 1998). Moreover, the FCRs used here in the calculations of waste loads are greater than those achieved by efficient farmers. For example, in whiteleg shrimp farming, some producers obtain FCRs of 1.1–1.3 (Boyd et al. 2021a, 2021b).

Dissolved oxygen demand

The dissolved oxygen demand in a culture system increases with greater feed input because the organic carbon in the feed that is not converted to harvest biomass can potentially be oxidized within the system, and most of the oxidation is realized within the system through the respiration of culture animals. The part of the organic carbon in the feed that is not eaten or that is ingested but expelled as feces is at least partially oxidized by the pond microbial community.

Aerobic respiration, whether by the culture species or microbial organisms, is realized through a complex series of biochemical reactions, but its overall stoichiometry is simple: (10) $$\text{Organic C}+{\mathrm{O}}_2\to {\mathrm{\text{CO}}}_2.$$(10)

The mass balances are: 2.67 kg oxygen to oxidize 1 kg organic carbon; 3.676 kg CO₂ released for each 1 kg or organic carbon oxidized; 1 kg oxygen consumed in aerobic respiration results in 1.38 kg carbon dioxide released.

The carbon waste load allows the carbonaceous oxygen demand associated with feed to be calculated by the relationship: (11) $$\mathrm{\text{CBOD}}={\mathrm{C}}_{\mathrm{L}}\times 2.67$$(11) Where CBOD = carbonaceous biochemical oxygen demand from feeding (kg O₂/t production), C_L = carbon waste load (kg/t production), 2.67 = O₂/organic C ratio.

The nitrogenous waste from feeds also causes an oxygen demand because oxygen is necessary for the biological oxidization of ammonia nitrogen by nitrifying microorganisms. The summary equation is: (12) $$\mathrm{N}{\mathrm{H}}_4^{+}+2{\mathrm{O}}_2\to \mathrm{N}{\mathrm{O}}_3^{-}+2{\mathrm{H}}^{+}+{\mathrm{H}}_2\mathrm{O}.$$(12)

The mass balance is two molecules of oxygen for each nitrogen atom oxidized. Thus, 4.57 g of oxygen are needed in the oxidation of 1.0 g of nitrogen in the ammonia or ammonium combination.

The nitrogen applied in feed is mostly in the form of amino acids and protein, and the major nitrogenous waste of aquatic animals is ammonia. Uneaten feed and feces will be decomposed by microorganisms with the release of ammonia. The ammonia nitrogen entering the culture system can potentially be oxidized to nitrate nitrogen. The nitrogenous biological oxygen demand resulting from ammonia nitrogen can be estimated from the nitrogen waste load as follows: (13) $$\mathrm{\text{NBOD}}={\mathrm{N}}_{\mathrm{L}}\times 4.57$$(13) where NBOD = nitrogen oxygen demand of feeding (kg O₂/t production), N_L = nitrogenous waste load (kg/t production), 4.57 = O₂/N in nitrification.

The estimated total BOD (CBOD + NBOD) resulting from the use of feeds at typical FCRs for the several species and expressed mostly within culture systems was calculated (Table 11). The estimated total BOD values for feeding ranged from 1365 kg O₂/t production for Pangasius to 3694 kg O₂/t for ictalurid catfish. The NBOD values were between 202 kg O₂/t and 912 kg O₂/t and averaged 18% (range = 14.3%–24.7%) of the total BOD. If ictalurid catfish are taken as a separate case and excluded, the range is constricted greatly and the averages and standard deviations for the other six species are: NBOD, 270 ± 49 kg O₂/t; CBOD, 1326 ± 139 kg O₂/t; total BOD, 1596 ± 169 kg O₂/t.

Table 11. Estimated carbonaceous, nitrogenous, and total biological oxygen demands (CBOD, NBOD, and BOD) and the aquatic acidification potentials (AAP) imposed by feeding for selected aquaculture species at typical feed conversion ratios (FCR).

Species	FCR	CBOD (kg O2/t)	NBOD (kg O2/t)	BOD (kg O2/t)	AAP (kg CaCO3/t)
Pangasius	1.5	1163	202	1365	315
Rainbow trout	1.3	1180	315	1495	492
Atlantic salmon	1.3	1292	216	1508	338
Whiteleg shrimp	1.6	1366	280	1646	438
Carp	1.7	1471	310	1781	484
Tilapia	1.7	1486	298	1784	466
Ictalurid catfish	2.5	2454	476	2930	745

As already mentioned, because most of the feed is eaten and the largest portion of the feed consumed is used in fish or shrimp as an energy source through respiration, the majority of the total BOD of feed will usually be expressed in shrimp or fish respiration within the culture system. Feed is made of highly digestible ingredients and uneaten feed and feces for the culture animals also will be decomposed rather quickly by heterotrophic microorganisms. A large amount of the organic waste from feed will be oxidized within most culture systems and especially in ponds (Boyd and Tucker 1998).

Environmental waste loads

Amounts of waste resulting from feed may not be entirely discharged into the environment; therefore, the environmental waste loads of most production facilities are less than the culture system waste loads. This is particularly the situation in ponds. Organic waste in ponds is oxidized by microbial activity and some of it accumulates in the bottom soil. Ammonia nitrogen from feeds is nitrified, and nitrate resulting from nitrification is denitrified by certain bacteria in anaerobic sediment (Hargreaves 1998; Gross et al. 2000). Dissolved phosphate is strongly adsorbed by sediment or precipitated as calcium phosphate directly from the water column (Hepher 1958; Masuda and Boyd 1994). Nitrogen and phosphorus also are contained in organic matter that accumulates in pond bottoms.

The efficiency with which carbon, nitrogen, and phosphorus in feed is converted into harvest biomass varies greatly with feed quality, feed management and resulting FCR, and environmental conditions in ponds. The efficiency with which the proportion of the feed entering ponds in waste varies with environmental conditions in ponds, especially with the adequacy of dissolved oxygen concentrations to support aerobic decomposition and nitrification, and retention time of water in ponds. The efficiency with which feed is converted to biomass and the feeding waste in ponds is assimilated before discharge is a critical issue with respect to water pollution. Some of the earliest studies on this topic were conducted for US ictalurid catfish farming. Based on several studies, Boyd and Queiroz (2001) concluded that about 5–10% of the carbon, 16–28.5% the nitrogen, and 7–11.4% of phosphorus added in feed were discharged in effluent. They also suggested that about 10 kg nitrogen and 1 kg phosphorus/1000 kg feed would be typical for ictalurid catfish, and they implied that this relationship likely would also apply to tilapia, and marine shrimp ponds.

There have been many subsequent studies on waste loads from aquaculture. In most of the studies, the discharges of nitrogen and phosphorus were measured, but inadequate information was provided to estimate the percentage of the feed input of nitrogen and phosphorus contained in the discharge. De Silva et al. (2010) conducted studies on Pangasius production in ponds using commercial feed in which nitrogen discharge ranged from 33.4–69.7 kg/t biomass and phosphorus discharge varied from 9.5–19.8 kg/t biomass. In ponds using farm-made feeds, the range was 1.9–83.4 kg/t and 6.0–36.2 kg/t for nitrogen and phosphorus, respectively. Some of the values exceed the system load for Pangasius for the commercial diets and FCR = 1.5 (Table 10). Of course, considerable water exchange is used in Pangasius farming and many farmers do not obtain a FCR as low as 1.5. These likely are the reasons for the high discharges of the elements found at some Pangasius farms by De Silva et al. (2010).

The study by DeSilva et al. also referenced some other species. Discharges from common carp ponds ranged from 30.9–86.0 kg nitrogen/t biomass and from 8.5–26.4 kg phosphorus/t biomass. The maximum values exceed the system loads estimated and reported in Table 10. Channel catfish farming in China discharges much more phosphorus than it does in the US. The ranges were 120–160 kg nitrogen/t biomass and 25–35 kg phosphorus/t biomass. The lower values exceed those calculated in the present study as typical system loads (Table 10), and this suggests that the FCR obtained for ictalurid catfish in China may be even greater than in the US.

Additional data from China (Zhang et al. 2015) are summarized in Table 12. The marine shrimp effluent discharge loads, when considered with system loads in Table 10 suggests that possibly 39.8% of nitrogen and 33.9% of the phosphorus in feed might have been discharged in the effluent. Using data for tilapia (Table 10) as the possible inputs of nitrogen and phosphorus for comparison to freshwater fish in Zhang et al.’s data, 23.3% of nitrogen and 15.8% of phosphorus in feed might have been discharged. The large standard deviations for the discharge quantities (Table 12) attest to the great variation in the amounts of nitrogen and phosphorus discharged from different ponds. These differences may have been related to several factors: feed quality, FCR, water exchange, and adequacy of the dissolved oxygen supply.

Table 12. Means and standard deviations for discharge of nitrogen and phosphorus in pond effluent in China.

	Number of ponds		Discharge (kg/t biomass)
	Nitrogen	Phosphorus	Nitrogen	Phosphorus
Mariculture
Fish	9	13	8.1 ± 7.8	3.0 ± 1.2
Shrimp	7	7	35.7 ± 17.8	5.9 ± 3.6
Freshwater
Fish	17	13	20.4 ± 32.1	3.2 ± 4.0
Shrimp	5	6	25.2 ± 17.5	6.1 ± 5.2
Crab	5	5	88.4 ± 176	11.1 ± 24.5

Source: Zhang et al. (2015).

The data summarized above show the great variability in the amount of nitrogen and phosphorus discharged from aquaculture ponds. The only reliable way of making this estimate is to measure the total discharge of effluent and to make enough analyses of the nitrogen and phosphorus concentrations in the discharge to obtain a reliable average. The appropriate equation is: (14) $$L_{\text{N or}\hspace{0.17em}\text{P}}=\hspace{0.17em}\frac{\text{V}\hspace{0.17em}\times \hspace{0.17em}{C}_{\text{N or}\hspace{0.17em}\text{P}}\hspace{0.17em}\times {10}^{-3}\hspace{0.17em}}{P}$$(14) where L = load of N or P (kg/t biomass), V = total effluent volume (m³), C = average concentration of N or P (g/m³, same as mg/L), 10⁻³ = t/kg, P = pond production (t biomass/crop).

In raceways and other flow-through units, there is insufficient water retention time for biological processes to remove appreciable amounts of nutrients and decompose organic matter. The discharge of flow-through systems will be dilute, but the nutrient loads may be large; loads of 47–71 kg nitrogen and 6.5–24.2 kg phosphorus per tonne of biomass have been reported (De Silva et al. 2010). MacMillan et al. (2003) attributed a 40% reduction in phosphorus from trout raceways to improving feeding practices and screening fish from tails of raceways to allow a quiescent area for sedimentation. Soderberg (2007) reported that 20% of total solids loads in flow-through systems could be removed in quiescent zones. Much of the nitrogen in culture system waters is in the form of ammonia nitrogen that cannot be removed as effectively as phosphorus. Foy and Rosell (1991) were able to recover a maximum of 9% of nitrogenous waste through sedimentation. Bergheim and Brinker (2003) were able to remove 3% and 21% of nitrogen and phosphorus from aquaculture effluent by screening.

In RAS fish and shrimp facilities, water treatments such as screening, sedimentation, and biofiltration are used to remove nitrogen, phosphorus, and suspended solids. According to Bregnballe (2015), screening can remove 15–27% total nitrogen, 50–84% total phosphorus, and 50–91% total suspended solids. Arellano (2020) concluded that 85–98% of organic matter and suspended solids and 65–96% of total phosphorus could be removed in RAS. There are claims of more efficient removal, but the estimates above are likely typical for commercial systems. Water must be discharged in RAS systems and make-up water added to avoid excessive total dissolved solid concentrations through evaporation. Such wastewater is concentrated in potential pollutants, and often is released into municipal sewage systems. In RAS and in intensive pond culture, some farms discharge wastewater into natural or constructed wetlands. According to Tom et al. (2021) wetlands can remove 86–99% of total nitrogen, 95–98% of total phosphorus, 25–35% of chemical oxygen demand, and 47–86% of total suspended solids.

In cage culture, all of the waste from feed goes directly into the water body in which cages are sited. According to Belle and Nash (2008), systems for collecting solid waste from cages have proved both impractical and too expensive to install and operate.

The use of water exchange in ponds shortens the time that water remains in ponds to lessen the amount of natural waste assimilation possible. In flow-through systems, the removal of solid waste by sedimentation also depends upon the flow rate and turbulence of water within culture units.

The only way to ascertain the actual amounts of carbon, nitrogen, and phosphorus added in feed that are discharged from most culture systems is by measurement of the loads of these elements in effluent. By lessening the FCR, system loads will be reduced, and a decrease in environmental loads will also result. The importance of FCR in reducing the efficiency of resource use and water pollution should not be overlooked.

Aquatic acidification potential

Nitrification of ammonia nitrogen from feed causes acidification (Equation 12(12) $$\mathrm{N}{\mathrm{H}}_4^{+}+2{\mathrm{O}}_2\to \mathrm{N}{\mathrm{O}}_3^{-}+2{\mathrm{H}}^{+}+{\mathrm{H}}_2\mathrm{O}.$$(12) ) at an equivalence of 7.14 kg CaCO₃/kg ammonia nitrogen (0.1428 kg H⁺/kg N). The potential acidity from feeding expressed in terms of equivalent calcium carbonate is: (15) $$\mathrm{\text{AAP}}={\mathrm{N}}_{\mathrm{L}}\times 7.14.$$(15)

The acidification potential can also be expressed in terms of hydrogen ion (16) $$\mathrm{\text{AAP}}={\mathrm{N}}_{\mathrm{L}}\times 0.1428$$(16) where AAP = potential aquatic acidity from feeding (kg CaCO₃/t biomass), N_L = system nitrogen load (kg/t biomass), 7.14 = CaCO₃/N ratio, 0.1428 = H⁺/N ratio.

The calculated values for the acidity that can result from feeding for the seven species (Table 11) ranged from 315 kg/t Pangasius to 745 kg CaCO₃/t ictalurid catfish. Omitting ictalurid catfish, the average and standard deviation were 422 ± 77 kg CaCO₃/t biomass.

The acidity from feeding will neutralize alkalinity, and the alkalinity may be restored by the dissolution of carbonate containing minerals in the bottom soil or by alkalinity of incoming water. In many situations, liming is necessary to avoid a decline in alkalinity. It is common practice to express alkalinity as its equivalent concentration of calcium carbonate. This allows the reaction of acidity to neutralize alkalinity to be taken as: (17) $$\mathrm{\text{CaC}}{\mathrm{O}}_3+2{\mathrm{H}}^{+}\to \mathrm{C}{\mathrm{a}}^{2+}+{\mathrm{H}}_2\mathrm{O}+\mathrm{C}{\mathrm{O}}_2\uparrow$$(17)

It is convenient to express both the aquatic acidification potential and the neutralizing value of liming materials (Table 13) in calcium carbonate equivalents as done also for alkalinity.

Table 13. Typical neutralizing values for selected liming materials used in aquaculture.^a

Liming material	Neutralizing value (kg CaCO3/kg)
Sodium bicarbonate	0.59
Agricultural limestone (calcitic)^b	1.00
Agricultural limestone (dolomitic)	1.09
Hydrated (slacked or builder’s) lime	1.35
Burnt (unslaked or quick) lime	1.79

^aThese are for the highest quality materials. Some vendors will have an exact analysis of the neutralizing values for specific liming products. Neutralizing value also may be given as a percentage (100 × values listed above).

^bPure calcium carbonate (CaCO₃) is used as the standard for assessing acid neutralizing capacities of specific liming material products.

The amount of alkalinity potentially neutralized as a result of feeding is identical to the acidification potential. The amount of liming material needed to counteract the acidification potential will depend upon the neutralizing value of the liming material used. To illustrate, the acidification potential of feeding tilapia is 466 kg/t biomass (Table 11). Pure calcitic limestone has a neutralizing value of 1.00 (Table 13), and 466 kg/t of this liming material would be needed. Using high quality burnt lime as the liming material, only 260 kg would be required [(466 kg CaCO₃/t tilapia) ÷ (1.79 kg CaCO₃ equivalent/kg CaO)].

The ratio of liming material as CaCO₃ equivalent at the FCRs and feed crude protein concentrations assumed in Table 10 ranged from 0.21 kg CaCO₃/kg feed for Pangasius to 0.38 kg CaCO₃/kg feed for rainbow trout with a mean and standard deviation for the seven feeds of 0.28 ± 0.05 kg CO₂/kg feed.

Atmospheric emissions

Carbon dioxide emissions occur through the respiration of farmed animals (MacLeod et al. 2020; USDA 2020), and of course, the decomposition of uneaten feed and feces of farmed animals results in the release of carbon dioxide by heterotrophic microorganisms. The carbon dioxide released directly as the result of feeding can be calculated from carbon waste loads that are reported in Table 10 because each kilogram of organic carbon oxidized yields 3.676 kg CO₂ as can be calculated from the stoichiometry of Equation 10(10) $$\text{Organic C}+{\mathrm{O}}_2\to {\mathrm{\text{CO}}}_2.$$(10) . The equation for calculating the carbon dioxide emissions from feeding is: (18) $${\mathrm{F}}_{\mathrm{d}}\mathrm{C}{\mathrm{O}}_2={\mathrm{C}}_{\mathrm{L}}\times 3.676$$(18) where F_dCO₂ = direct carbon dioxide emission from feeding (kg CO₂/t biomass), C_L = system carbon load (kg organic carbon/t biomass), 3.676 = kg CO₂/kg organic C.

The estimated direct carbon dioxide emissions for the seven species ranged from 1603 kg CO₂/t for Pangasius to 3378 CO₂ kg/t for ictalurid catfish (Table 14). Omitting ictalurid catfish as an outlier, the average and standard deviation for the other species is 1827 ± 192 kg CO₂/t.

Table 14. The atmospheric emissions as CO₂ and CO₂e from feeding on seven common aquaculture species.

	Direct CO2 (kg/t)		Embodied CO2e (kg/t)
Species	Feed	Alkalinity depletion	Feed	Liming materials	Total (CO2e/t)
Carp	2025	213	425	44	2707
Ictalurid catfish	3378	328	725	68	4499
Tilapia	2048	205	540	42	2835
Pangasius	1603	139	460	29	2231
Atlantic salmon	1779	149	1,302	31	3261
Rainbow trout	1625	216	1,334	45	3220
Whiteleg shrimp	1882	193	654	40	2769

The reaction of the potential acidity from the nitrogen load from the feed with alkalinity in the water also releases carbon dioxide directly (Equation 17(17) $$\mathrm{\text{CaC}}{\mathrm{O}}_3+2{\mathrm{H}}^{+}\to \mathrm{C}{\mathrm{a}}^{2+}+{\mathrm{H}}_2\mathrm{O}+\mathrm{C}{\mathrm{O}}_2\uparrow$$(17) ). From the stoichiometry of Equation 17(17) $$\mathrm{\text{CaC}}{\mathrm{O}}_3+2{\mathrm{H}}^{+}\to \mathrm{C}{\mathrm{a}}^{2+}+{\mathrm{H}}_2\mathrm{O}+\mathrm{C}{\mathrm{O}}_2\uparrow$$(17) , each kilogram of alkalinity (as CaCO₃) neutralized by acidity yields 0.44 kg carbon dioxide. The potential aquatic acidity values (Table 11) can be used to estimate the amount of carbon dioxide released: (19) $$A_{{\text{CO}}_2}=\text{AAP}\times 0.44$$(19) where $A_{{\text{CO}}_2}$ = carbon dioxide from neutralization of alkalinity (kg CO₂/t biomass), 0.44 = CO₂/CaCO₃ ratio.

The amounts of carbon dioxide from neutralizing alkalinity (Table 14) ranged from 139 kg/t for Pangasius to 328 kg/t for ictalurid catfish. These are rather small in comparison from the direct emissions from feeding.

Methane and nitrous oxide in aquaculture

There is a question with respect to atmospheric emissions that should not be ignored in pond aquaculture. How much of the highly potent GHGs, methane and nitrous oxide, are emitted by bottom soils and sediments in pond aquaculture? Much of the concern about GHGs in aquaculture has focused on carbon dioxide, but there are other GHGs to include methane, nitrous oxide, hydrofluoro-carbons, and sulfur hexafluorides. The fluorine containing GHGs are commonly called F-gases. Methane has a global warming potential 28-fold greater than carbon dioxide and nitrous oxide is 289 times as potent as carbon dioxide. Carbon dioxide emissions account for 74.4% of the total global warming potential of anthropogenic GHG emissions. The percentages for the other gases are 17.3% for methane, 6.2% for nitrous oxide, and 2.1% for F-gases (Ritchie et al. 2020).

The biggest fraction of global, farm production of freshwater fish and crustaceans, milkfish and other estuarine fish species, penaeid shrimp, and crabs results from pond culture (Boyd et al. 2022). Pond bottoms consist of the original soil upon which solids continually settle. Pond soil and sediment is waterlogged during aquaculture crops and becomes anaerobic. The anaerobic zone begins a few millimeters or at most 1–2 cm below the sediment-water interface (Boyd et al. 2010; Munsiri et al. 1995). In the anaerobic zone of pond bottoms, certain species of anaerobic bacteria convert carbon dioxide to methane and nitrate and nitrite to nitrous oxide.

The onset of anaerobic conditions in pond bottoms first results in fermentation, a type of anerobic respiration in which carbon dioxide, alcohols, fatty acids, acetates, ketones and other organic components are released into the water in the pores of bottom soil and sediment (Boyd 2020). Organic fermentation products can be used in respiration by anaerobic, chemotrophic bacteria. The oxygen in certain inorganic ions and compounds serves as terminal electron and hydrogen ion acceptors instead of dissolved oxygen in the anaerobic decomposition of fermentation products by chemotrophic bacteria.

Denitrification usually is the first inorganic chemotrophic process to occur in pond bottoms after the redox potential has declined below that necessary for fermentation to begin (Boyd 1995, 2020). The basic reactions of interest in denitrification follow: (20) $$\mathrm{\text{HN}}{\mathrm{O}}_3+2{\mathrm{H}}^{+}\to \mathrm{H}{\mathrm{\text{NO}}}_2+{\mathrm{H}}_2\mathrm{O}$$(20) (21) $$2\mathrm{H}\mathrm{N}{\mathrm{O}}_2+4{\mathrm{H}}^{+}\to {\mathrm{H}}_2{\mathrm{N}}_2{\mathrm{O}}_2+2{\mathrm{H}}_2\mathrm{O}$$(21) (22) $${\mathrm{H}}_2{\mathrm{N}}_2{\mathrm{O}}_2+4{\mathrm{H}}^{+}\to 2\mathrm{H}{\mathrm{N}}_2\mathrm{O}\mathrm{H}$$(22) (23) $$\mathrm{N}{\mathrm{H}}_2\mathrm{O}\mathrm{H}+2{\mathrm{H}}^{+}\to \mathrm{N}{\mathrm{H}}_3+2{\mathrm{H}}_2\mathrm{O}$$(23) (24) $${\mathrm{H}}_2{\mathrm{N}}_2{\mathrm{O}}_2\to {\mathrm{N}}_2\mathrm{O}+{\mathrm{H}}_2\mathrm{O}$$(24) (25) $${\mathrm{H}}_2{\mathrm{N}}_2{\mathrm{O}}_2+2{\mathrm{H}}^{+}\to {\mathrm{N}}_2+2{\mathrm{H}}_2\mathrm{O}$$(25) (26) $${\mathrm{N}}_2\mathrm{O}+2{\mathrm{H}}^{+}\to {\mathrm{N}}_2+{\mathrm{H}}_2\mathrm{O}$$(26)

In denitrification, most of the nitrate and nitrite are ultimately reduced to elemental nitrogen gas (N₂), but ammonia and particularly nitrous oxide also results and sometimes in large amounts depending upon conditions (Boyd 1995).

The presence of nitrate and nitrite maintains a constant redox potential, but when these two forms of inorganic nitrogen have been used up in denitrification, the redox potential will fall and manganic, ferric, and sulfate compounds will be used progressively as the redox potential continues to decline in chemotrophic respiration by other types of organisms. Finally, the redox potential falls to the level necessary for carbon dioxide to be used by methane forming bacteria. Methanogenesis typically is explained in equation form as follows: (27) $$\mathrm{C}{\mathrm{H}}_3\mathrm{\text{COOH}}+2{\mathrm{H}}_2\mathrm{O}\to 2\mathrm{C}{\mathrm{O}}_2+8{\mathrm{H}}^{+}$$(27) (28) $$8H^{+}+C{O}_2\to C{H}_4+2H_{2}O$$(28)

The sequence of events that occurs in pond bottoms is typically stratified with a depth of the sediment. Aerobic microbial decomposition occurs in the thin oxidized sediment layer just below the sediment-water interface. Within the sediment below, there will be a progression of sediment layers in which the different types of chemotrophic bacterial processes occur in response to a declining redox potential that diminishes with greater sediment depth (Boyd 2020).

Sediment depth was measured in cores of pond bottoms from 233 ponds of known age in nine countries (Boyd et al. 2010). The sedimentation rate averaged about 1.0 cm/yr. The average carbon sequestration rate in sediment was 1490 kg organic carbon/ha/yr equivalent to 5460 kg CO₂/yr. These authors commented that anaerobic sediment was a source of methane and nitrous oxide emissions that would likely offset the benefit of carbon sequestration, but to an extent not known at that time. Interest in methane and nitrous oxide emissions has since increased, and some measurements of the emission rates of these two GHGs from pond bottoms have been made.

Methane

The majority of the studies of methane and nitrous oxide emissions from aquaculture ponds have been conducted in China, and Dong et al. (2023) recently reviewed the Chinese literature on methane emissions from aquaculture ponds. Data were available from 55 experiments involving 37 ponds. The average for all the studies was 6.68 mg CH₄/m²/hr (66.8 g/ha/hr) with a 95% confidence interval of 20 g CH₄/ha/hr. The highest emission was for a shrimp pond (247.4 g CH₄/ha/hr) and the lowest for a carp polyculture pond (0.66 g CH₄/ha/hr). The averages for ponds with mixed species (any combination of fish, shrimp, and crabs), crab ponds, and shrimp ponds were 24 g CH₄/ha/hr, 74 g CH₄/ha/hr, and 127 g CH₄/ha/hr, respectively.

Considerable variation was noted in the estimates of methane emissions and there was no statistical difference (p > 0.05) between the averages for mariculture ponds (84 g CH₄/ha/hr) and freshwater ponds (58 g CH₄/ha/hr). Aeration was noted to reduce methane fluxes from freshwater ponds, and water exchange lessened emissions from mariculture ponds.

Emissions from polyculture ponds (various combinations of fish species and crab) resulted in less than 0.25 kg CH₄/kg protein in edible meat, while the methane to protein ratio in shrimp ponds was 1–2 kg CH₄/kg protein. Aquaculture protein was less methane intensive than the methane to protein ratio of 3–4 kg CH₄/kg protein for beef cattle, buffalo, sheep, and goat meat protein.

Dong et al. (2023) estimated that aquaculture ponds in China released 1.6 Mt/yr of methane into the atmosphere. This was compared to 0.2 Mt/yr of methane from reservoirs in China, 11.25 Mt of methane from lakes, natural wetlands, and rice paddies, and a country-level total emissions of 61.5 Mt CH₄/yr of methane. The conclusion was that aquaculture ponds contributed about 2.6% of national emissions in China.

Cage culture also can be a source of increased methane and nitrous oxide emissions from lakes and reservoirs. Pu et al. (2022) reported methane emission of 48 g CH₄//m²/yr, but a lake not impacted by cage culture had methane emission of 8 g CH₄/m²/yr. Cage culture apparently was included in the estimate of total methane emissions by Chinese aquaculture.

Carp ponds in the Czech Republic were reported to emit an average of 3.2 g CH₄/ha/hr (Rutegwa et al. 2019) while Znachor et al. (2023) reported a range of 6–9 g CH₄/ha/hr. Integrated rice-fish ponds in India had average methane emissions of 92.7 kg/ha over a 168-day period or 23.0 g CH₄/ha/hr (Datta et al. 2009). Odinga et al. (2023) gave average methane emissions from extensive tilapia ponds in Kenya as follows: no fertilizer, 0.13 g/ha/hr; inorganic fertilizer, 0.3 g/ha/hr; organic fertilizer, 0.08 g/ha/hr.

The aquaculture area in China, excluding the surface area of cages, was reported to be 6.56 M ha in 2019 (Hu et al. 2021) and an area of 7.0 M ha was given by Textor (2022). Using the 7.0 M ha and the 1.6 Mt/ha estimate of methane by Dong et al. (2023), the average methane emission by ponds in China is 229 kg CH₄/ha/yr. A study by Malerba et al. (2020) of small agricultural ponds (used for farm purposes and not for aquaculture) released methane at an average of 258 kg/ha/yr. This study focused on the United States with 2.56 million agriculture ponds with an area of 420,900 ha (average pond area of 0.19 ha) and Australia with 1.76 million agriculture ponds covering 291,200 ha (average pond area of 0.17 ha). This study gives a similar estimate of methane emission per water surface area as the one calculated from the data presented by Dong et al. (2023) for aquaculture ponds in China. The important issue raised by the study of small agricultural ponds relates to whether methane production in ponds is greatly influenced by pond use. The main factor may be the nature of the sediment in anaerobic zones of pond bottoms rather than the specific use of the ponds.

Munsiri et al. (1995) and Boyd (1995) noted that bottom soils tended to reach an equilibrium concentration of organic carbon within a period of 2–10 years. The management of ponds might only effect methane emissions to the extent that they influence the equilibrium concentration of bottom soil organic carbon. The work by Boyd et al. (2010) did reveal that low input aquaculture ponds usually had slightly lower organic carbon concentrations than did feed-based production, but considering all data collected, most ponds had 1–3% soil organic carbon.

Assuming a methane emission rate of about 250 kg/ha/yr, the carbon dioxide equivalent of the methane would be about 7000 kg CO_2e/ha/yr. This exceeds the average carbon assimilation capacity of pond bottoms of about 5460 kg CO₂/ha reported by Boyd et al. (2010). Aquaculture pond bottom soils must be considered net emitters of GHGs based on the typical emission rates of methane alone.

Nitrous oxide

Less information was found on nitrous oxide emissions of aquaculture ponds. Ma et al. (2018) reported that a pond stocked with black carp, grass carp, and chub emitted an estimated 4.7 kg N₂O/ha/yr while a crab pond had an emission of 5.2 kg N₂O/ha/yr. The hourly emission estimates were highly variable ranging from near 1 to about 250 µg N₂O/m²/hr. Fang et al. (2022) measured nitrous oxide emissions over 2 years in crab ponds and in carp polyculture ponds. Annual nitrous oxide emissions averaged over three ponds for each type of aquaculture and both years were 1.83 kg N₂O/ha/yr for fish ponds and 2.53 kg N₂O/ha/yr for crab ponds. The average flux of nitrous oxide was given as 28.5 µg N₂O/m²/hr.

The average of the estimates of nitrous oxide emission given above is 3.6 kg/ha/yr. The average nitrous oxide emission of 3.6 kg/ha/yr is equivalent to 1040 kg CO_2e/ha/yr. This is roughly 15% of the CO_2e estimated methane emissions from Chinese studies. Aquaculture pond emissions of methane and nitrous oxide together may average around 8040 kg CO_2e/ha, but some portion of this would be counteracted by carbon sequestration in pond bottoms.

Ponds release carbon dioxide from sediment also, but the study by Boyd et al. (2010) suggests that more carbon dioxide is sequestered than released. The implication is that aquaculture pond bottoms contribute to global warming mainly by emitting methane and nitrous oxide. Aeration apparently reduces methane and nitrous oxide emissions from ponds (Dong et al. 2023; Fang et al. 2022), but the use of feed in aquaculture is a source of organic sediment in pond bottom soils which increases the amounts of organic carbon and nitrogen and increase the potential for methanogenesis and denitrification. Organic and commercial fertilizers increase phytoplankton production in ponds and dead plankton is an organic matter input to pond bottoms. The nitrogen in fertilizers also can stimulate denitrification. Extensive ponds may also be important sources of aquaculture GHGs.

Aquaculture pond soils are a minor global source of methane and nitrous oxide emissions, but they must be considered in efforts to lessen global GHG emissions. There is no practical means known for oxidizing pond bottom soils below oxidized surface layers (Boyd and Tucker 1998). The main methods for reducing methane and nitrous oxide in aquaculture ponds are to apply no more organic or commercial, nitrogen fertilizer than necessary, to strive to obtain a better FCR in feed-based aquaculture, and to provide adequate mechanical aeration in intensive production.

The methane and nitrous oxide data were reported on a pond area basis. In China, the production of fish and other species, excluding cages, was 39.0 t in 2019 (Hu et al. 2021). Based on a 7 M ha area this is a production of 5.57 t/ha. The average methane flux was calculated above to be 229 kg/ha/yr which equates to 89.1 kg CH₄/t biomass or 2228 kg CO_2e/t biomass. The average bottom soil sequestration rate (Boyd et al. 2010), might offset 980 kg CO₂/t biomass. The average amount of nitrous oxide emitted per hectare of 1040 kg CO_2e/ha/yr equates to 187 kg CO_2e/t biomass/yr. Based on these averages, the expectation is about 1435 kg CO_2e from aquaculture production in systems with earthen bottoms.

The database for methane and nitrous oxide emissions from aquaculture is rather small and limited primarily to China. There is evidence that some systems emit more methane than others, but the factors causing the differences have not been carefully investigated. There is even less data on nitrous oxide emissions than for methane emissions. It is clear that methane and nitrous oxide are emitted from aquaculture ponds with earthen bottoms, and further investigation is warranted.

Embodied GHG emissions

Feed also has embodied atmospheric emissions associated with the production, processing, and delivery of feed ingredients to feed mills, milling of feed, and delivery to farms. The GHG emissions resulting from the production of grid electricity and the production and use of hydrocarbon fuels and include carbon dioxide, methane (CH₄), and nitrous oxide (N₂O).

The global warming potential of methane and nitrous oxide are 25 and 289 times greater (weight basis), respectively than that of carbon dioxide (Bender and Davis 2012). Carbon dioxide accounts for about 81% of total global GHG emissions. Most sources of carbon dioxide emissions other than aerobic respiration and neutralization of acidity have methane and nitrous oxide associated with them. The common practice is to calculate the carbon dioxide equivalents of methane and nitrous oxide emissions from a particular source, convert them to equivalent amounts of carbon dioxide, and add the carbon dioxide equivalents of the methane and nitrous oxide emissions to the direct carbon dioxide emissions from a source. The resulting estimate is called the carbon dioxide equivalent (CO_2e) of GHG emissions from a source. The CH₄ and N₂O emission factors for diesel fuel and gasoline (USEIA 2022) were adjusted to CO₂e factors and added to the CO₂ emission factor per liter of each fuel to provide a CO₂e factor (kg CO₂e/L) for each fuel (Table 15).

Table 15. Direct carbon dioxide emission factors for selected aquaculture pond inputs and a general CO₂e factor for embodied emissions in production of inputs.

Item	CO2e	CO2
Electricity (kg CO2e/kW·hr)	0.647
Diesel fuel (kg CO2e/L)	2.763
Gasoline (kg CO2e/L)	2.329
Molasses (kg CO2/kg		0.73
Sugar (kg CO2/kg)		1.47
Liming materials (kg CO2/kg):
Agricultural limestone, calcitic (CaCO3)		0.44
Agricultural limestone, dolomitic (CaCO3·MgCO3)		0.48
Burnt lime (CaO)		0.78
Hydrated lime [Ca(OH)2]		0.60
Sodium bicarbonate (NaHCO3)		0.54
Embodied in inputs (kg CO2e/GJ embodied energy/t harvest biomass	70.56

There is much information on the CO₂e factor for electricity in which methane and nitrous oxide have been included. However, the CO₂e of electricity is problematic in estimating the GHG contribution of electricity at a particular location. Electricity is generated by several methods, and CO₂e ranges from 0.013 kg/kW·hr for nuclear generation and wind generation to 0.486 kg/kW·hr for generation from natural gas to 0.840 kg/kW·hr for generation from oil and 1.001 kg/kW·hr for coal (NREL 2021). The electricity in grid networks often consists of electricity from mixtures of electricity from two or more methods of generation. The world average CO₂e for grid electricity is 0.515 kg/kW·hr (Carbon Footprint 2019). The average CO₂e for grid electricity in the five major aquaculture exporting countries follow: India, 0.743 kg/kW·hr and Indonesia, 0.755 kg/kW·hr (Carbon Footprint 2019); Vietnam, 0.913 kg/kW·hr (Department of Climate Change 2019); Thailand, 0.424 kg/kW·hr (CEIC 2022); Ecuador, 0.400 kg/kW·hr (Parra 2020). For purposes of the present study, the five-country average of 0.647 kg/kW·hr will be used (Table 15). For comparison, CO₂e of the US electrical grid averages 0.385 kg/kW·hr varying from 0.106 kg/kW·hr (UPCC New York grid) to 0.709 kg/kW·hr (HICC Oahu grid) in Hawaii (USEIA 2021). In the EU, the average CO₂e is around 0.225 kg/kW·hr but varies from 0.025 to 0.70 kg/kW·hr among the 25 member nations (EEA 2022).

Data could not be found for calculating directly the amounts of embodied CO₂e associated with each feed, but embodied energy data were available (Table 5). The global amounts of the major industrial energy sources, electricity, diesel fuel, gasoline, liquid petroleum gas (LPG), and natural gas, were obtained (BP 2022). The CO₂e factors for major fuels (SEAI 2017) were weighted according to the amounts of each used to provide a weighted, general CO₂e factor of 70.56 kg CO₂e/GJ of embodied energy. Using this factor, the embodied CO₂e associated with feeding can be calculated as: (29) $${\mathrm{F}}_{\mathrm{e}}\mathrm{C}{\mathrm{O}}_{2\mathrm{e}}=\mathrm{\text{FEE}}\times 70.56\times \mathrm{\text{FCR}}$$(29) where F_eCO₂e = embodied CO₂e potential of feed (kg CO₂e/t feed), FEE = embodied energy content of feed (GJ/t feed), 70.56 = CO₂e factor (kCO₂e/GJ embodied energy).

Applying Equation 29(29) $${\mathrm{F}}_{\mathrm{e}}\mathrm{C}{\mathrm{O}}_{2\mathrm{e}}=\mathrm{\text{FEE}}\times 70.56\times \mathrm{\text{FCR}}$$(29) , the CO₂e emissions based on the embodied energy contents of generic, commercial diets (Table 5) were computed. The embodied CO₂e values for the seven feeds ranged from 425 kg CO₂e/t for Pangasius to 1334 kg CO₂e/t for rainbow trout (Table 13).

The embodied CO₂e in liming material to restore alkalinity was estimated from the embodied energy in agricultural limestone 1.29 GJ/t. The quantities ranged from 29 kg CO₂e/t for Pangasius to 68 kg CO₂/t for ictalurid catfish (Table 14). The totals of the direct CO₂ and embodied CO₂e emissions ranged from 2231 kg CO₂e/t for Pangasius to 4499 kg CO₂e/t for ictalurid catfish ($\overline{x}$ ± SD = 3,075 ± 717 kg CO₂e/t).

Atmospheric acidification potential

The energy sources used in making feed resulted in the release of acid-forming gases into the atmosphere: nitrous oxide (N₂O), sulfur dioxide (SO₂), and ammonia (NH₃) which typically are referred to as NO_x, SO_x, and NH_x as the exact composition is questionable. Sulfuric acid (H₂SO₄) results from oxidation of SO₂ and nitric acid (HNO₃) results from oxidation of NO₂ and NH₃. Complete oxidation of 1 mole each of NO₂ and NH₃ yield 1 mole H⁺, 1 mole of SO₂ yields 2 mole H⁺. Thus, 1 g NO₂ is equivalent to 0.696 g SO₂ while 1 g NH₄ is equivalent to 1.778 g SO₂. These factors were used to convert NO₂ and NH₄ emission factors (Haneke 2002) to an SO₂ basis, and these were added to the amount of SO₂ emitted to provide the SO₂ equivalent (SO₂e) factors for diesel fuel and gasoline (Table 16). The SO₂e for electricity (Table 16) was taken from Nugroho et al. (2022). The general embodied SO₂e factor of the other inputs (Table 16) was derived from the weighted average of the SO₂e factors (see above) for five common industrial fuels in the same manner that the CO₂e factor for inputs was derived above earlier.

Table 16. Sulfur dioxide equivalence factors of acidic atmospheric emissions of electricity, diesel fuel, and gasoline.

Source	Factor
Electricity^a (g SO2e/kW·hr)	5.89 g SO2e/kW·hr
Diesel^b (g SO2e/L)	6.68 g SO2e/L
Gasoline^b (g SO2e/L)	5.48 g SO2e/L
General factor for embodied SO2e in inputs^b (g SO2e/GJ/t shrimp)	334 g SO2e/GJ/t shrimp

^aNugroho et al. (2022).

^bSee calculation details in the text.

The embodied energy in the feeds and liming materials used to replace alkalinity were used with the assumed FCRs and the general SO₂e factor of 334 g SO₂e/GJ to estimate the atmospheric acidification potential for feeding (Table 17). The values for SO₂e had the ranges: feed, 2.01–6.31 kg SO₂e/t; liming material, 0.14–0.32 kg SO₂e/t; total, 2.22–6.52 kg SO₂e/t.

Table 17. Atmospheric acidification potential (kg SO₂e/t biomass) related to feeding and replacement of depleted alkalinity in feed-based culture for seven aquacultural species.

	kg SO2e/t biomass
Species	Feed	Liming material	Total
Pangasius	2.18	0.14	2.32
Rainbow trout	6.31	0.21	6.52
Atlantic salmon	6.16	0.15	6.31
Whiteleg shrimp	3.09	0.19	3.28
Carp	2.04	0.21	2.22
Tilapia	2.56	0.20	2.76
Ictalurid catfish	3.43	0.32	3.75

Aeration and atmospheric emissions

Energy for aeration could be estimated only for white leg shrimp and ictalurid catfish because no reliable estimates were found for typical amounts of aeration energy use in farming other species. The estimates, assuming the use of electric aerators, were 10 GJ/t for ictalurid catfish and 11.9 GJ/t for white leg shrimp, the equivalent of 2778 and 3306 kW·hr/t, respectively. These amounts of electricity would result in 1797 kg CO₂e/t ictalurid catfish and 2139 kg CO₂e/t for shrimp. The sulfur dioxide emissions from electricity use were calculated as 16.4 kg SO_2e/t for ictalurid catfish and 19.5 kg SO₂e/t for white leg shrimp.

Reference to Table 14 gives direct and embodied carbon dioxide emissions directly related to feed range from 2063 kg CO₂e/t for Pangasius to 4103 kg CO₂e/t for ictalurid catfish, respectively. The average and standard deviation was 2,826 ± 655 kg CO₂e/t. Of course, the carbon dioxide emissions associated with the use of liming materials were small in comparison to feed and aeration. They averaged 8% of the total carbon dioxide emissions of 3075 kg CO₂e/t. The carbon dioxide emissions for electricity use in aeration are 70.9% and 80.8% of the total emissions for ictalurid catfish and shrimp, respectively.

The sulfur dioxide emissions associated with electricity use for aeration were estimated as 16.4 kg SO₂e/t ictalurid catfish and 19.5 kg SO₂e/t whiteleg shrimp. Reference to Table 17 shows that this is about 4.8 times more sulfur dioxide than resulted from feed alone for ictalurid catfish. The corresponding difference for whiteleg shrimp is 6.3 times more carbon dioxide. Of course, liming material was the source of comparatively little sulfur dioxide.

Contribution of carbohydrate sources

Molasses, sugar, or some other carbohydrate sources are commonly used in biofloc technology for the culture of shrimp and some fish species. The use of carbohydrate additions also has been adopted in both semi-intensive and intensive pond culture of shrimp and fish, especially shrimp by some farmers. For example, sugar was applied at an average rate of about 1000 kg/ha/yr at 81 farms in a sample of 101 shrimp farms in Ecuador, none of which used the biofloc method (Boyd et al. 2021b). In Thailand, 12 farms in a sample of 34 farms that were not applying biofloc technology used either sugar or molasses at an average of 1151 kg/ha/yr (Boyd et al. 2017).

The benefit of carbohydrate sources in non-biofloc culture systems is not clear. In biofloc methodology, the carbohydrate is used by heterotrophic microorganisms for growth resulting in the conversion of ammonia excreted by the culture species and microbial decomposition of feeding waste into microbial protein (Avnimelech 2015). Removal of the ammonia is a benefit because ammonia is potentially toxic to shrimp and fish. The microbial floc that develops in biofloc systems can be used by shrimp and some fish species for food. Biofloc technology potentially allows more efficient use of feed protein through nitrogen recycling. It is likely that such a benefit would result in ordinary feed-based culture, but no studies to verify this possibility were found.

In Ecuador, the average molasses use was 143 kg/t shrimp. The land, water, and energy use per tonne of shrimp resulting from molasses use was calculated as 0.003 ha, 75 m³ freshwater, and 0.07 GJ, respectively. Had sugar been used instead of molasses, the respective values would be 0.012 ha, 238 m³, and 0.27 GJ. The use of molasses, which is a coproduct of sugar production from sugarcane, requires less resources than does the use of raw sugar.

The direct CO₂ emissions from blackstrap molasses (about 20% carbon) and raw sugar (about 40% carbon) would be 105 kg CO₂/t shrimp and 210 kg/t shrimp, respectively. The embodied energy contents of molasses and sugar would be 5 kg CO_2e/t shrimp and 19 kg CO_2e/t shrimp, much less than the direct carbon dioxide. While much lower than for feed, carbohydrate sources can result in carbon dioxide emissions. This example with sugar and molasses shows again how the selection of an input can influence the amounts of impacts that can potentially be expressed.

Relationship to LCA

The approach used here to estimate resource use and negative impacts resulting from the use of feeds in aquaculture production is in many ways similar to that of life cycle assessment (LCA). The application of LCA is currently popular for assessing the negative impacts of many aspects of production activities. It is widely used in agriculture (Clark and Tilman 2017) and it is growing in popularity in aquaculture (Bohnes et al. 2019; Bohnes and Laurent 2019). One of the benefits of utilizing LCA is that modern computers can store, process and run calculations on many variables and very large data sets. Life cycle assessment is typically accomplished by inputting data for a production system into software programs, that have been made with specific assumptions. According to Renouf et al. (2018), 14 software programs have been customized for conducting agricultural product LCAs, but none were found that were specifically adapted to aquaculture products.

The use of LCA software will be most reliable where the software and the underlying assumptions are closely designed for the specific inputs made to the product system which is under investigation. If this is accomplished, the power of modern computer processing would generate a large amount of information on the minor and major aspects of the products’ life cycle. If this is not the case, the investigator(s) will have to make appropriate modifications. The output of an LCA would then be able to reveal the major areas where efficiency gains can be made, and perhaps as important, what are the minor factors amounting to rounding errors and would not have an impact on magnitude, such as lime and molasses used at some feed-based aquaculture facilities.

An experience will be related in regards to the previous paragraph. The authors conducted an LCA by direct calculation (not with LCA software) for shrimp aquaculture based on primary data previously collected from 2016 to 2020. This was a tedious task, and when accomplished, the results were sent to a group of investigators who were interested in developing aquaculture-specific LCA software. Because of the need to identify both the major and minor impacts to assess the life cycle of farmed shrimp, we included all aspects of facility construction and maintenance and all embodied resources to make the effort as inclusive as possible. The response from the LCA experts was astounding: the authors’ effort for a complete LCA was considered too detailed and specific for what the LCA experts said that they would consider in an LCA. With this response, we recognized the serious shortcomings of an over-reliance on software and an under-reliance on primary data collection. Moreover, the published literature describing aquaculture LCAs often were incomplete by omitting various inputs and embodied energy associated with inputs.

The initial results of LCAs usually are midpoint indicators, global warming potential, freshwater acidification potential, freshwater pollution potential, marine pollution potential, land occupation, freshwater consumption, fossil fuel use, and others. Most LCA software calculates 14–18 midpoint indicators (Huijbregt et al. 2017). These indicators are similar to the potential resource use and impacts presented in this review for the potential amounts of pollutants, effects that result from the production system or the aspect of the production system under consideration. In the case of the present assessment of aquaculture feed, the way the calculations were made is obvious and the required equations are presented. The user of LCA software may or may not know or even attempt to determine the calculation procedures (and resource use and other coefficients) that were used by the software developers.

In LCA, the software goes beyond the initial calculations of midpoint impacts. The estimated midpoint impacts are multiplied by damage factors and aggregated into categories, usually human health, ecosystems, and natural resource use (Matlock et al. 2022). Many product systems have one or more aspects that can negatively affect humans. Some software generates what is known as the disability adjusted life years (DALY) of an individual or a community or an individual exposed to the negative health effect impacts of the LCA target product (Scanlon et al. 2013). The findings of LCA are highly speculative and especially when projecting beyond the calculation of the midpoint indicators.

The present assessment of aquaculture feed and earlier assessments of the farm-level aspects of whiteleg and black tiger shrimp production (see review of four studies by Boyd et al. 2021a) demonstrate the highly variable nature of producer performance. For the informed, this variability is natural and common, but for those who seek singular numerical values to describe the impacts of a production system without recognition of the variability, LCA results can be misunderstood or misreported. The amount of variability influences the outputs of LCAs, and it is important to understanding the context and precision of targeted estimates. Two LCA approaches can be applied to a single aquaculture product or feed ingredient and result in a difference in magnitude. The complexity of LCA methods coupled with some of the unitless indices provided suggest that we should continue to find new ways to simplify and report on issues where there is great variability. The approach illustrated in this report is an alternative to LCA, and it also is a procedure that requires primary data collection. While LCA probably is adequate for a general assessment, a resource use assessment such as the assessment of feed given here can give more specific data on resource use and for potential, negative environmental impacts. Of course, in the present study, the results were based on what was perceived as the typical practices. The same approach could be applied to specific situations of any particular aquaculture production facility.

While this assessment is restricted to feed and adequate data to make a more complete assessment of resource use for the species were not the goal, it is clear from data on shrimp farms (Boyd et al. 2017; Boyd et al. 2021a, 2021b) that the greatest share of energy use is related to feed use and mechanical aeration. Shrimp farming was, based on protein yield as edible meat, similar to broiler chicken and pig production less than beef cattle production with respect to land and water use. Shrimp production required more energy than used for any one of these traditional meats. It is suspected that this same pattern could be expected for any type of feed-based aquaculture in which mechanical aeration is used.

The requirement for mechanical aeration is related to the fact that natural sources of dissolved oxygen are not adequate to supply the amount of molecular oxygen that is necessary to support the aerobic respiration of the culture animals and the aerobic respiration of heterotrophic microorganisms that oxidize the waste from feed and the oxygen needed in biological nitrification. Boyd et al. (2018) discuss the disadvantage of aquatic animals to terrestrial animals with respect to respiration. The calculations made by Boyd et al. (2018) revealed that in order to move 1 g of molecular oxygen across respiratory surfaces require 4.8 g air for terrestrial animals and 110 kg water for fish and shrimp.

The situation with aquatic respiration is further conflicted by the fact that the farmed animals are living in the water and the combined respiration of all aerobic organisms in removing the scant supply of oxygen from the water. Freshwater at 25 °C and standard pressure contains only 8.25 g molecular oxygen per m³—approximately 8.25 g/t or 0.00825% of molecular oxygen. While the air is about 21% molecular oxygen, it enters the water at a much slower rate than the aquaculture and other aerobic organisms consume molecular oxygen (Boyd et al. 2018). In the daytime in ponds, aquatic plants do produce a large amount of oxygen, but at night photosynthesis while respiration continues. Nighttime presents the most critical concern over dissolved oxygen depletion.

There really is no practical means of avoiding the need for aeration in most aquaculture systems with high feeding rates. By adopting better feeds and feeding practices, the amount of oxygen demand associated with feed can be reduced. Better aerators and aeration practices can lessen energy use per unit of aquaculture production. No suggestions other than these will be offered, but there obviously is a need to improve feed-based, aerated aquaculture to lessen energy use and GHG emissions through feasigle and cost effective practices.

Conclusions

Considerable amounts of land, energy, water, and wild fish associated with feed manufacturing become embodied in fish and shrimp biomass produced in feed-based aquaculture. Mechanical aeration applied to allow high feeding rates also resulted in a large input of energy in electricity or hydrocarbon fuels necessary in intensive production of aquatic animals. Much nitrogen, phosphorus, and organic waste are associated with feeding, and energy use and the oxidation of the feed by the culture animals and the microbial decomposers of feeding waste result in atmospheric emissions. Lesser quantities of emissions result from liming material that may be added to neutralize acidity from feed and the degradation of carbohydrate sources applied to stimulate biofloc production in high intensity culture of shrimp and other species.

There are two main ways of reducing resource use and aquatic pollution and associated atmospheric emissions from feed. One is to design feed formulae that require smaller inputs of embodied land, water, energy, and wild fish. It probably is not possible to lessen all four resources as the results of the tradeoffs, e.g. less fishmeal leads to less wild fish use, but more land use. Nevertheless, some combinations of ingredients can lower one or more of the four major resources of concern.

The other way is to improve feed management on farms to lessen the FCR. There likely is more opportunity to improve the environmental performance of feed-based aquaculture at the farm level than at the feed mill level. Of course, where aeration is applied, the use of efficient aerators and regulation of aerator use in harmony with the actual amount of aeration required can lower energy use and atmospheric emissions associated with this practice.

A large amount of feed-based aquaculture is conducted in ponds that have anaerobic soils from which methane and nitrous oxide, two gases with a greater global warming potential than carbon dioxide, are emitted. Pond sediments also release carbon dioxide, but carbon sequestration in sediment is greater than carbon dioxide emissions. The emissions of methane and nitrous oxide appear great enough to more than offset the benefit of carbon sequestration. Complete studies of GHG emissions from pond aquaculture must include a consideration of net carbon sequestration and emissions of methane and nitrous oxide from sediment.

Data obtained for this review point to a broader challenge facing the aquaculture sector and perhaps the whole food sector. The deviations from applied research to more novel research for high-impact factor publications appears to have diminished the continuous supply of primary data describing the status of aquaculture in specific countries and regions. Additionally, the aquaculture sector may be more reticent to share data because they could be challenged for the impacts the data reveal. Data are being collected and we know there are very large amounts of producer data being housed by feed companies and perhaps processing facilities. There will always be value in holding information for a competitive advantage. Those who control the information can shape the perception of the industry. There is a danger that reliance on the virtual and modeled world could be or could become grossly inaccurate without our knowledge. As natural resource scarcity increases, accuracy of the estimates of resource use will become more important.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

This research was supported by a grant from the Gordon and Betty Moore Foundation to the World Wildlife Fund.