Z. Aquafeed, marine ingredients and carbon feedstocks to 2050

What this appendix contains. The evidence and calculations behind Protein balance 2050 (sections 17.5 and 17.6) and Frontier technology (section 16.5). Part 1 (Z.1 to Z.8) covers targets, our aquaculture paths to 2050, trash fish, offshore cages, marine ingredients, omega-3, protein crops and climate. Part 2 (Z.9 to Z.14) asks whether residues, side streams and one-carbon feedstocks could supply the glucose that the S-ALT scenario needs, and what released rice land could grow. Technology status and costs of cellulosic sugar and one-carbon routes are in V. Frontier technology; the balance model is in W. Balance model.

How to use it. Feed manufacturers and investors: Z.3, Z.4 and Z.6. Policy makers: Z.1, Z.3, Z.10 and Z.12. Founders choosing a fermentation site: Z.9, Z.11 and Z.13. Signposts are in X. Drivers and signals.

---

Part 1. Aquafeed and marine ingredients

#Z.1 Official targets and published projections

• No Vietnam-specific projection beyond 2035. We checked Decisions 339, 1664 and 389 and Fish to 2050 and found no published volume after 2035 and no national species-level target for 2030 ^4,9 VN-directHigh. The only later anchor is the 2045 vision that marine aquaculture should supply "trên 25% tổng sản lượng" (over 25% of total output), read as total fisheries output ⁵ VN-directMedium.
• The published model sits below the target and ignores the fleet cut. OECD-FAO holds capture near 3.5 Mt in 2030, 25% above the 2.8 Mt target (our subtraction of aquaculture from total fish) ^8,3 VN-directMedium projection. Its 2025 base (5.69 Mt) is below the NSO figure of 6.12 Mt ¹¹ VN-directHigh. We show both and do not average them.
• Marine aquaculture is behind plan: about 730 kt in 2021 ¹² against 850 kt targeted for 2025; we found no 2025 outturn VN-directMedium.
• Lobster value is ahead of the volume plan. Lobster exports to China were USD 845 M in 2025 and over USD 506 M in January to May 2026 (+44.3%) ¹³, against a 2030 output target of 5 kt VN-directMedium signal.
• Growth after 2030 is the largest open number. IMPACT implies ASEAN aquaculture growth of about 0.5% a year from 2030 to 2050 (our calculation from ⁹) VN-adjacentLow projection, while Vietnamese aquaculture grew 5.1% in 2025 and 5.7% in the first half of 2026 ^11,14 VN-directHigh trend.

#Z.2 Our three aquaculture paths

All paths: VN-directLow estimate. Marine finfish is carved out of "other fish" from an assumed 80 kt in 2025.

Output by species group, kt live weight. 2025 measured ¹¹ VN-directHigh; later years our calculation, aqf_calc.py VN-directLow estimate.

Feed and protein needs (our calculation, aqf_calc.py) VN-directLow estimate

Coefficients (our assumptions). FCR on pellets 1.5 for all groups except whiteleg shrimp (1.3), improving 0.3% a year. Dietary CP: pangasius 27%, shrimp 40%, other fish 30%, marine finfish 45%; the shrimp and marine values sit within the 40 to 43% of standard shrimp feeds ¹⁵ and the 40 to 50% of marine feeds ¹⁶ VN-directMedium. Share of output on pellets from 2025 to 2050: pangasius and whiteleg 100%; other shrimp 30 to 50%; other fish 68 to 85%; marine finfish 15% rising to 60% (LOW), 95% (BASE) or 100% (HIGH).

How AQ-BASE relates to S-BASE. AQ-BASE gives 8.86 Mt of aquaculture and 9.72 Mt of aquafeed in 2050, against 8.52 Mt and 9.24 Mt in S-BASE (balance_outputs.csv, BLO-0223, BLO-0235). The gap is only the faster marine finfish path; aquaculture growth after 2035 moves the 2050 soybean-meal result by about 1 Mt at most (W. Balance model) VN-directLow estimate. Across our paths, 2050 aquafeed protein spans 1.6 Mt (2.5 to 4.1 Mt of CP): growth after 2030 matters more than formulation.

#Z.3 Trash fish: use, supply cuts and replacement

Cá tạp (trash fish: low-value fish, crustaceans and molluscs, mostly trawl by-catch).

Pangasius and whiteleg shrimp have since moved to pellets, and up to 0.3 Mt of snakehead pellets in 2025 suggests most snakehead has too ² VN-directLow.

The capture plan. Decision 389 caps trawlers at about 8,360 by 2030, about 59% fewer than in 2016 (our calculation from ^6,18) VN-directMedium projection, and capture is to fall from 3.83 Mt (2025) ¹¹ to 2.8 Mt (2030) ³ VN-directHigh projection. Trawl catch is the main source of trash fish, so low-value fish for direct feeding and for whole-fish meal should fall sharply by 2030 VN-directMedium estimate. Decision 339 orders "thức ăn công nghiệp thay thế sử dụng cá tạp" (industrial feed to replace trash fish) and Decision 1664 orders feed zones next to sea farms, but neither sets a volume or date ^3,5 VN-directHigh.

Direct trash-fish feeding in 2025 (our calculation, aqf_calc.py) VN-directLow

The central case is about 20% of 2025 capture (10 to 42%) and carries about 133 kt of protein at 17% CP. Pellets that replace it need only about 88 kt of CP (54 to 144 kt), because they convert about four times better. The 2002 figure is not comparable: the species mix has changed.

Extra formulated feed over 2025, from marine finfish growth plus conversion of the other trash-fish species (15, 30 and 50% converted by 2030 in LOW, BASE and HIGH; 40, 70 and 90% by 2040; 60, 90 and 100% by 2050; our calculation, aqf_calc.py) VN-directLow estimate

Fishmeal is at 20% of marine pellets in 2025, falling to 12% by 2050 (the hypothesis in Feed and aquafeed market, H6, is 15 to 30%). Reading: to 2050 the transition is a small share of aquafeed tonnage but a large share of the premium segment, where a novel protein or oil competes with fishmeal and fish oil, not soybean meal VN-directLow estimate.

#Z.4 Marine finfish and offshore cages

Decision 1664 targets for fed species ⁵ VN-directHigh projection (official target): marine finfish 120 kt in 2025 and 200 kt in 2030 (of which offshore 60 and 120 kt); lobster 3 and 5 kt; other crustaceans 67 and 95 kt; all fed species 190 and 300 kt (our sum). Offshore (nuôi biển xa bờ: offshore sea farming) is 340 kt of all species on 30,000 ha by 2030. The offshore provinces are, on the current map, Quang Ninh, Hai Phong, Quang Ngai, Dak Lak (former Phu Yen), Khanh Hoa (including former Ninh Thuan), Lam Dong (former Binh Thuan), Ho Chi Minh City (former Ba Ria-Vung Tau), Ca Mau and An Giang (former Kien Giang).

The Khanh Hoa pilot. Decision 231/QD-TTg (January 2025) allows 240 ha at 0 to 3 nautical miles (3,600 t) and 200 ha at 3 to 6 nautical miles (5,100 t) to 2029 in HDPE cages, phased 30 ha (2025), 100 ha (2026 to 2027) and 110 ha (2028 to 2029) ⁷ VN-directMedium projection. The phases add up to 240 ha; we could not see how the outer 200 ha is phased. Pilot HDPE farms reported profits of 172% (cobia), 112% (lobster) and 131.4% (grouper) of same-size wooden-cage farms (provincial claim) ⁷ VN-directLow. Storm No. 12 (2017) destroyed wooden cages there ²³, and HDPE cages hold up to 300 t of cobia each ²¹ VN-directMedium.

Feed capacity. Marine-fish feed capacity was 40,000 to 50,000 t a year in 2022, over 80% foreign-owned ²² VN-directMedium. De Heus opened a marine and cold-water fish feed mill in Vinh Long in July 2026, 168,000 t a year by the company's figure ¹⁶ VN-directMedium signal. We found no national count or target for recirculating (RAS) farms.

Where protein and omega-3 concentrate (marine finfish pellets; our calculation, aqf_calc.py) VN-directLow estimate

Offshore growth is pellet growth, because offshore cages cannot use trash fish at scale. The 120 kt offshore target for 2030 alone would need about 177 kt of pellets (120 kt x 1.5 x 0.985), twice AQ-BASE's 2030 estimate (our calculation) VN-directLow estimate.

#Z.5 World marine ingredients to 2035 and 2050

• OECD-FAO builds El Niño years into 2027 and 2031, with Peru's fishmeal at 0.72 and 0.53 Mt against about 1.1 Mt in normal years ⁸ generalMedium projection.
• 2026 runs below the model. Peru's first 2026 season landed about 25% of a 1.9 Mt quota, and world fishmeal output fell 26% year on year in January to April 2026 ^25,26 generalMedium. NOAA gives a greater than 90% chance of a very strong El Niño in winter 2026 to 2027 ²⁷ generalHigh signal.
• Our 2050 extension: whole-fish meal flat or down at 3.2 to 3.8 Mt, plus by-product meal of 2.8 to 3.8 Mt, up from about 1.9 Mt in 2024 (aqf_calc.py) generalLow estimate. Later this century, extreme El Niño events about double in frequency under high emissions in older models ²⁸, a finding contested for the newest ones; the IPCC finds ENSO rainfall variability very likely amplified in 2050 to 2100 ²⁹ generalLow projection. Almost 12 Mt of processing by-products are still not collected, most of the potential in Asia ²⁴ generalMedium.

Vietnam's own supply

• Pangasius off-cuts are the base. Pangasius supplied 12% of the world's by-product fishmeal and 29% of its by-product fish oil in 2024 ²⁴ VN-directMedium: about 225 kt of meal (5.52 Mt x 34% x 12%) and 215 kt of oil (1.375 Mt x 54% x 29%), or 0.116 t of meal and 0.111 t of oil per t of fish (our calculation). By-products are 62 to 67% of the whole fish ³¹ VN-directMedium.
• The 2050 fishmeal-demand range runs from 151 kt (S-EFF, S-ALT) and 253 kt (S-BASE) to 393 kt if the OECD-FAO 2035 value grows 1% a year (our calculation) VN-directLow estimate.
• Reading. Vietnam can plausibly stay self-sufficient in fishmeal volume to 2050 through by-products, but stays short of high-grade (65% CP) meal for shrimp and marine feeds, which it imports today (Feed and aquafeed market, H5) VN-directLow estimate.

#Z.6 Omega-3: needs, supply gap and alternatives

EPA plus DHA need in aquafeed, kt a year (central values; our calculation, aqf_calc.py) VN-directLow estimate

The ranges use 0.3 to 1.0% for shrimp, 0.7 to 1.5% for marine finfish ³² and 0 to 0.2% for freshwater fish. AQ-LOW gives 14.8 kt in 2050. The AQ-BASE 2050 need equals about 85 kt of marine fish oil at 25% EPA plus DHA (41 to 159 kt), part of it supplied by residual oil in fishmeal.

Vietnam's fish oil is not an omega-3 source. Oil from pangasius processing waste holds 0.07 to 0.15% EPA and 0.10 to 0.16% DHA in its fatty acids, in a study of Indonesian farmed pangasius ³⁰ VN-adjacentMedium. Vietnam's 190 kt of fish oil therefore holds only about 0.4 to 0.6 kt of EPA plus DHA (our calculation), against a need of about 11 kt VN-directLow.

The global gap. World EPA plus DHA supply is "optimistically estimated at just over 0.8 million tonnes", with a shortfall of more than 0.4 Mt, or over 1 Mt in pessimistic calculations; almost 90% comes from capture fisheries ³³ generalMedium.

Alternatives on the market

Credible 2050 shares for alternative ingredients (AQ-BASE; our calculation, aqf_calc.py) VN-directLow estimate

• Why omega-3 can go further than protein: Vietnam has no domestic EPA plus DHA; the global gap is structural; the products are commercial and trialled in shrimp; and value per tonne of feed is high VN-directLow. On protein, the soybean-meal slice is priced out at 2026 costs (4.6 to 5.1 times soybean-meal protein, Feed and aquafeed market, H11), and Chinese single-cell protein competes for the fishmeal slice.
• Binding conditions to 2050: the fish-oil price in El Niño years, listing of algal and transgenic oils as aquafeed raw materials, and import cost VN-directLow estimate. A Vietnamese algal-oil plant would need sugar or glycerol, which links omega-3 to Part 2.
• Chapter 19 adopts the 15 to 50% range as a normative goal (Vision 2050).

#Z.7 Protein crops and biotech rules

Soybean in Vietnam

• We use 20 to 39 kha for current area and OECD-FAO for projections; the yield series disagree and we leave that unresolved VN-directMedium. Domestic beans cost VND 25,000 to 30,000 per kg against VND 13,000 to 15,000 for imports, and profit is about VND 20 million per ha ³⁹ VN-directMedium. Domestic beans go to food, whose use alone was 540 kt in 2025 ⁴⁰ VN-directMedium.
• The crop strategy (Decision 1748) sets no soybean, maize or cassava target ⁴¹ VN-directMedium. A press proposal for winter soybean on 300 kha of Red River Delta rice land is not an adopted plan ³⁹ VN-directLow.

What-if: domestic soybean in 2050. Bean-equivalent need is soybean-meal need divided by 0.78: 13.3 Mt in S-BASE and 9.0 Mt in S-EFF (our calculation) VN-directLow estimate.

Domestic beans would first displace food-soy imports, so their contribution to feed protein by 2050 is close to zero unless better varieties close the yield gap and a feed-grade price is supported VN-directLow estimate.

Biotech rules

Reading. Gene editing is the only long-run route by which domestic soybean could matter for feed; the first signals to watch between 2026 and 2035 are a notification under Decree 43/2026 or a field-test application VN-directMedium signal. Duckweed and azolla have been fed in northern Vietnam for centuries ^43,44, but we found no current area or volume data VN-directMedium.

#Z.8 Climate and aquaculture geography

About 80% of Vietnamese shrimp is farmed in the Mekong Delta ⁴⁹ VN-directLow. We found no dated SSP-based projection of Vietnamese shrimp or pangasius output, and no quantified link between heat and disease. On current evidence, climate to 2050 changes where aquafeed is used more than how much VN-directMedium estimate.

---

Part 2. Carbon feedstocks

#Z.9 The S-ALT glucose need and competing uses

The sugar route in S-ALT needs 238 kt of glucose in 2030, 852 kt in 2040 and 1,318 kt in 2050 (balance_outputs.csv, BLO-0888, BLO-1048, BLO-1128) VN-directLow estimate. S-ALT is a stretch above every real-world analogue (Protein balance 2050), so this is an upper test, not an expected demand.

Source: balance model, tools/balance_model.py VN-directLow estimate.

Cassava-starch glucose costs about USD 511 to 560 per t at April 2026 prices (our estimate from starch at USD 520 to 540 FOB ⁵⁷, 1.08 t of glucose per t of starch, plus USD 30 to 60 for hydrolysis) VN-directLow. The nitrogen need (0.37 t of urea per t of protein) does not change with the carbon source. The question for Part 2 is how much carbon could come from residues, side streams and one-carbon feedstocks without cutting food or starch exports.

#Z.10 Rice straw

Rơm rạ (rice straw and stubble).

• Straw yield is about 4.72 t per ha, at 12.4% moisture when baled ⁶⁴ VN-directMedium; straw is about 38% cellulose, 25% hemicellulose and 12% lignin, with high ash ⁶⁵ generalMedium.
• About half the national straw is surplus to current uses, 55% of it in the Mekong Delta; former Kien Giang alone could run 245 MW of straw power (2019 basis) ⁵⁹ VN-directMedium.
• The national figure is robust to about 20%: straw-to-grain ratios run 0.7 to 1.4 in Thailand ⁶⁶, and Mekong data imply 1.14 ⁶⁷ VN-directMedium.

Straw needed for the S-ALT sugar route at 0.31 to 0.44 t SE per t of air-dry straw (the low end for organisms that cannot use xylose; our calculation, ngf_calc.py) VN-directLow estimate

All national straw could in theory give 12.8 to 21.5 Mt SE in 2050, 10 to 16 times the need. Tonnage is not the constraint.

Competing uses

Costs. Mechanised collection costs USD 12 to 18 per t ⁶⁴ VN-directMedium; at USD 25 to 50 per t delivered, straw adds USD 57 to 162 per t SE for the feedstock alone (our calculation) VN-directLow. Technology in brief: mature-plant models give USD 342 to 467 per t of cellulosic sugar, but first-of-a-kind plants cost USD 2,100 to 2,400 per t of annual capacity, and India's rice-straw plant at Panipat reached 62% of design only in December 2025, after its first problem, buying straw from farmers who preferred to burn it ^{71,72,73,74,75} generalMedium. Details: V. Frontier technology.

#Z.11 Cassava pulp, liquid side streams, food waste and manure biogas

Technical sugar-equivalent potential if the whole stream were used in 2050 (our calculation, ngf_calc.py) VN-directLow estimate:

Bagasse (0.71 to 1.01 Mt SE), maize stover (1.40 to 2.00) and coffee pulp (0.07 to 0.14) are lignocellulosic and mostly committed to boilers, fodder or fertiliser (feedstock_futures.csv, FFU-003, FFU-006, FFU-007).

• Cassava pulp is the only near-term second-generation carbon of scale. Its starch needs only ordinary amylases, and 2024 volumes were about 60% of 2025's. Diverting pulp removes a feed carbohydrate, so the net gain is the protein made (Feedstock data sheets, C3) VN-adjacentLow.
• Liquid side streams are small but free and at a factory gate. Enifer's revival of the PEKILO mycoprotein process is the model to watch: a 3,000 t a year plant under construction in Finland and a 500 t pilot on corn-ethanol thin stillage in Brazil, both reported in 2025 ⁹³ generalMedium signal (company claims).
• Pig-manure methane (our estimate): 31.4 M pigs ⁹¹ x 0.3 kg of volatile solids per head a day x 0.29 m3 of methane per kg x 60% gives about 0.43 Mt a year, of which the large-farm share (30% in 2025, 65% by 2050, our assumption) is capturable VN-directLow estimate. We found no national biogas figure.
• One-carbon feedstocks in brief. Methanol-fed yeast protein is at pilot scale in China, at about 4.6 t of methanol per t of protein ⁹⁴ VN-adjacentMedium signal. The whole 2050 S-ALT need via e-methanol would take about 1.5 Mt of methanol, 0.29 Mt of green hydrogen (1.5 to 2.9% of the 2050 hydrogen target ⁹⁵), 2.1 Mt of CO2 and about 16 TWh (our calculation) VN-directLow estimate. The limit is cost, not volume (V. Frontier technology).

#Z.12 Rules as a ceiling

• EU feed law is precise. Regulation 767/2009, Annex III, bans from feed "All waste obtained from the various phases of the treatment of the urban, domestic and industrial waste water", "irrespective of any further processing of that waste", plus household waste and faeces. A footnote exempts process water in independent conduits that carries feed or food material and is free of cleaning agents ⁹⁶ generalHigh.
• Vietnamese rules are permissive but vague. We found no legal ban on food waste or wastewater-derived substrates for feed; the veterinary authority only advises cooking kitchen waste at 100 °C for 20 to 30 minutes before feeding pigs ⁹⁷ VN-directLow. The feed listing of microbial biomass is the real gate (Rules).

How each stream fares under EU-style rules (our reading of the Annex) generalMedium

Exporters that follow EU-style rules would reject protein grown on wastewater-treatment streams, food waste or manure-derived gas, whatever Vietnamese law allows, so from 2026 onwards this is a ceiling on such carbon, and Cases A and B send only small shares of these streams to protein VN-directMedium signal. A Vietnamese rule that follows the EU line between process water and treatment waste would lift the uncertainty (Robust moves, RM-06).

#Z.13 Cases A, B and C for 2040 and 2050

Case rules: share of each stream sent to fermentation in 2050, with 2040 in brackets where different (our assumptions, ngf_calc.py). No case cuts food or starch exports; released land is excluded. All VN-directLow estimate.

Results: Mt SE a year and share of the S-ALT sugar-route need (our calculation, ngf_calc.py) VN-directLow estimate

Case B in 2050, step by step (Mt SE): straw 1.0 Mt x 0.31 to 0.44 = 0.31 to 0.44; pulp starch 0.48 to 0.98 x 25% x 0.97 = 0.12 to 0.24; e-methanol 0.10; stover 0.04 to 0.06; stillage 0.03 to 0.05; manure methane 0.03; bagasse 0.02 to 0.03; starch wastewater 0.02 to 0.04; the other five streams 0.06 to 0.11. Total 0.73 to 1.11.

Plants and investment at first-of-a-kind costs of USD 2,100 to 3,400 per t of annual sugar capacity, from Panipat and Podari (our calculation) VN-directLow estimate

Meeting the whole S-ALT need from straw sugar would take USD 1.8 to 2.9 billion by 2040, and USD 2.8 to 4.5 billion and 15 to 21 Panipat-size plants by 2050. We did not cost side-stream plants or an e-methanol industry.

• Case A is today's reality extended: about a tenth of the 2050 need, almost all from cassava pulp. The carbohydrate ceiling stands.
• Case B turns the ceiling into a capital question: five Panipat-size plants (or two of 500 kt of straw) running reliably by 2050, plus side-stream plants at starch factories and ethanol plants. The trigger is an Asian straw plant running reliably above 80% of design (Robust moves, RM-19).
• Case C removes the ceiling but needs about 10% of national straw and an e-methanol industry: an ambitious bet, not a plan.
• No case changes the nitrogen need or makes the carbon low-carbon by default. One Indian bagasse-sugar study gives 1.57 kg CO2e per kg of sugar, above Thai cassava starch at 0.60 to 0.97 kg ^98,99 generalLow.
• Vietnam has no 2G pilot for straw, bagasse or coffee husk ^69,70 VN-directMedium. China listed 35 "non-grain bio-based" cases in January 2026, yet BBCA's non-grain lactic acid is 1,000 t within 500,000 t of capacity ¹⁰⁰ VN-adjacentMedium signal.

#Z.14 Land: released rice land and protein per hectare

The adjusted national land plan (Decision 1177/QD-TTg, June 2026) sets rice land at 3,252.63 kha in 2030 against 3,907.07 kha in 2024, a release of about 654 kha ⁶⁰ VN-directMedium projection (official plan). Against the 2021 plan, agricultural land falls by 1,120.62 kha and non-agricultural land rises by 1,192.43 kha ¹⁰¹ VN-directMedium.

Where it is going. In the 2026 winter-spring season rice area fell 37.3 kha. In the south, 12 kha went to other annual crops, 5.6 kha to perennial crops and 1.9 kha to aquaculture (Can Tho alone 14.8 kha); in the north, conversions went mostly to non-farm uses, such as 800 ha for an airport and 938 ha for industrial zones and transport in Bac Ninh ¹⁰² VN-directMedium trend. Under Decree 112/2024, land that switches crops, or combines rice with aquaculture ponds of up to 20% of the plot, still counts as rice land ¹⁰³ VN-directMedium; so the 654 kha is mainly land leaving agriculture. No plan names a protein crop ^60,41 VN-directMedium. If all 654 kha grew cassava for fermentation after 2030 it would give 3.3 Mt of glucose a year, an upper bound on first-generation carbon from soils that suit cassava poorly (our calculation) VN-directLow estimate.

Protein per hectare (all 15 rows of land_protein_yields.csv; our calculation unless stated)

• Per hectare, fermentation beats Vietnamese soybean: cassava and cane via microbes give about twice Vietnam's soybean protein and about the same as imported soybean meal, so cassava-based microbial protein moves land use rather than saving it VN-directLow.
• Straw adds protein with no extra land, but only if cellulosic sugar becomes bankable.
• On wet released land, the land-efficient options are aquatic and forage systems, not grain legumes. Duckweed (bèo) is already on the livestock feed list (Rules); we found no tropical field data for its yield VN-directLow.

#Z.15 Gaps, disagreements and open questions

Gaps (these feed Q. Open questions)

Disagreements (for R. Disagreements)

#Data files

• aquaculture_projections.csv (44 rows): 2025 output by species, official targets, published projections, and our paths, aquafeed, trash-fish and extra-pellet estimates.
• marine_ingredients_2050.csv (39 rows): world and Vietnamese fishmeal and fish oil to 2035 and 2050, by-product shares, pangasius supply, EPA plus DHA needs and credible alternative shares.
• protein_crops_vn.csv (29 rows): soybean and maize data, GE approvals, Decree 43/2026, soybean what-ifs, released rice land, duckweed.
• feedstock_futures.csv (17 rows): each residue, side stream and one-carbon route with volumes to 2050, sugar-equivalent yield, cost, TRL, competing uses and rules.
• land_protein_yields.csv (15 rows): protein per hectare by crop or system.
• balance_outputs.csv (1,150 rows): here, the S-ALT glucose, cassava, sugar and urea rows and S-BASE aquaculture and aquafeed.
• climate_impacts_2050.csv (34 rows): here, rows CI-11, CI-12, CI-24, CI-25 and CI-30.
• Scripts and notes: working-papers/wave3/aquaculture_futures/ and working-papers/wave3/next_gen_feedstocks/.

Related: Protein balance 2050, Frontier technology, V. Frontier technology, W. Balance model, Feed and aquafeed market, Feedstock data sheets, Vision 2050, Robust moves, X. Drivers and signals, Q. Open questions, R. Disagreements.