NEOBIOME

D02: Smart Food Systems & Agriculture

Properties1
typeedt

60 min read

D02: Smart Food Systems & Agriculture

Domain framework: edt_framework

Scope

On-site food production, processing, and storage using technology-enabled and regenerative methods. Grounded in IPCC AR6 Ch. 7 (AFOLU) and FEW Nexus literature. Deliberately separated from water systems (D03) to maintain distinct measurement frameworks — FSR for food, WSR for water. CR_004 CR_005

Key technologies

Permaculture systems, aquaponics, hydroponics, vertical farming, precision agriculture, soil health monitoring, food forests, AI-assisted crop management, composting systems, cold storage, seed saving systems, food processing and preservation.

Forest Lodge Orchard — electric cherry orchard (CR_034)

  • 9 ha, ~9,300 UFO high-density cherry trees (NZ Zero brand; Taiwan/Shanghai + an Auckland boutique chain at a ~15% premium) — the premium flips the electrification payback to <2 yr. The UFO planar cordon intercepts 20–30% more light → needs ~2× the water of a traditional orchard (50–65k L/ha/day peak) — the food→water coupling in the flesh (cf. CR_033). World’s-first 100% electric food production. See forest_lodge_orchard. CR_034

SSI connections

  • I02 Food security & sustainable agriculture — FSR = Production / (Production + Imports − Exports) × 100 is the primary metric. CR_002
  • I07 Fulfilment of basic needs — food is a fundamental basic need.
  • I09 Environmental sustainability — regenerative agriculture builds soil carbon and reduces chemical inputs.
  • I01 Financial & economic sufficiency — local food production reduces household food expenditure and import dependency.

Evidence

  • Grazing establishment CapEx completed. The stockyards and breeding-stock legs are now sourced: cattle yards ~15,000 (firm packages 9,900-14,100 URL_038); breeding stock treated as one-off CapEx (project design decision) at IRD national-average values OT_222, breeding cow ~1,500/head and ewe ~165-180 long-run (smoothing the documented 2026 saleyard peak). With OT_212 / OT_213 (fence + water reticulation), the pasture-grazing establishment CapEx is fully sourced.
  • Dairy CapEx + crop-water (cost + irrigation gaps). House-cow dairy CapEx ~15,000-35,000 for self-supply (portable milker 3,106.55 URL_037; a refrigerated vat + MPI registration $135 are CONDITIONAL, only if milk is sold REG_035; the herd-scale milking plant stays price-on-application). Crop water requirement per crop = FAO-56 Kc x NZ ETo (OT_218): potato ~550 mm (NZ field trial 491-611 LIT_130), faba ~400, hemp/oilseed ~450 (rapeseed proxy), field veg ~400; effective-rain fraction 0.75 (USDA-SCS) vs ~0.6-0.65 (FAO OT_219), best a daily balance (OT_220 applied demand 170-400 mm/yr; 80% efficiency, Sep-Apr OT_221).
  • Farm-machinery energy (the D05 diesel-vs-electric lever). Cropping diesel ~100 L/ha/yr (FAR direct NZ litres, OT_215, down from the interim 130); grazing splits into dairy ~50 and sheep & beef ~15 L/ha (crude derivation: OT_126 off-road diesel volumes / NZ land areas RD_035 + RD_036); horticulture ~100 L/ha low-confidence placeholder. Electrification factor ~1.3 kWh per litre displaced on a same-task basis (Monarch MK-V spec URL_036, down from the conservative 3.0). Off-road diesel ~$2.00/L (MBIE RD_034; NZ diesel carries no excise, so the pump price approximates off-road).
  • Farm-machinery CAPITAL. Diesel baseline: a NZ compact tractor + front-end loader lands ~25,700 (23 HP) to 51,150 (49 HP) ex-GST (Massey Ferguson RRP, URL_039); implements add ~$5-20k. Electric sensitivity: Monarch MK-V starting ~USD 88,999 (~NZD 150k, ~3x diesel; price-on-application, URL_036). Equipment inventory (what a small community food supply needs, two tiers, walk-behind + compact tractor): URL_040.
  • Livestock water + local feed-grain land. Dairy stock water ~70 L/cow/day drinking + ~50 shed washdown = ~120 L/cow/day (beef 45, ewes ~3), Horizons RC guideline OT_225, closes the audit’s asymmetric coupling (the shed’s electricity was modelled, its water was not). Local hen-feed land: NZ feed barley ~7 t/ha (OT_226, USDA GAIN reporting FAR) -> ~64 m2/hen at 45 kg feed/hen/yr, so local-feed eggs draw arable land instead of being SSI-free.
  • Solar-agricultural coupling: Aftrak’s model uses agricultural productivity (food income) as the economic entry point for solar investment, then extends to community electricity. Dr. Wilson: “We need to address the lack of revenues. We need to address the lack of food, and then we can address the lack of electricity.” Validates an integrated D02/D01 design sequence for communities where food productivity and energy access are co-dependent constraints. Journalistic source. URL_001
  • Deep bed farming (DBF) + electric mechanisation: solar micro electric tractor eliminates manual hardpan-breaking; pierces compacted soil to increase root depth, water absorption, and drought resilience; reduces field preparation from 3–4 days to 1–2 days. Tiyeni claims 2× crop yield and up to 9× farmer revenue (unverified). URL_001
  • Dual-use modular solar: 7.5 kWh/day per module — charges one tractor per day or powers up to 5 houses; modules connect “like Lego” to scale with community size. Cooperative and hourly leasing (“Uber-like”) business models under development. URL_001
  • Agri-PV (dual food+energy land use, Fraunhofer ISE): growing crops and generating solar on the same land raises land-use efficiency ~60–80% (land-equivalent ratio ~1.6–1.8) — the technology that lets a land-constrained community avoid trading food area for energy area; two system families (ground-level <2.1 m vs elevated overhead), partial shade can buffer crops against heat/drought. German guide — NZ cost/consent a gap (RT_108). See agro_pv. OT_050
  • Retained cropland after an agri-PV build, and a reality-check on the +60–80% land-use-efficiency headline (Reher et al. 2024, Applied Energy, Belgium). Field-measured footprint land loss is 11% (near-ground vertical/tracked) to 8% (elevated) → ~0.89–0.92 of the field remains in crop production, the parcel-conservation input for the food-vs-energy land lever (kept separate from the under-panel yield derate). Measured whole-system LER was only 1.00–1.22 (0.97 in one wheat year) for open-field arable in a cool maritime climate — materially below the theoretical ~1.6–1.8 (+60–80%) carried from OT_050, so treat that headline as best-case, not a planning expectation for temperate arable staples. ⚠ Belgium proxy, no NZ value. LIT_114
  • Under-panel food yield — the pasture/grazing derate (Andrew et al. 2021, Frontiers, Oregon). Herbage under agri-PV runs 9–33% below open pasture (whole-system ~0.85–0.95 of open-field; fully-shaded strips worst at −38%, partially-shaded ≈ open), while lamb liveweight is comparable (1.5 vs 1.3 kg/ha/day, P=0.67) — lower forage quantity offset by higher quality. Grazing is the best-evidenced, lowest-derate pairing (order: pasture > shade-tolerant horticulture > arable staples). Sets the multiplier on grazing/veg yield when co-sited with PV. ⚠ Oregon proxy (44°N temperate); the sign flips with climate — partial shade can help in hot/dry, hurt in cool/cloudy — so model as ~0.85–0.95 nationally with an irradiance/aridity modifier, not a flat number, and transfer the ratio, not the levels. LIT_115
  • NZ vegetable nitrogen fertiliser rates (Reid & Morton 2019 / OT_069; compiled in CR_056). Crop-specific soil-test-adjusted total-seasonal fertiliser-N: brassicas 0–230, leafy greens 0–120, roots to 0–240, onion 0–140 kg N/ha — 0 at high soil Available N (fertiliser N = crop demand − soil supply), so these are NOT crop N uptake. The verified NZ re-source for the model’s veg N-demand after LIT_108 (Tei 2020) was found not to support CROP_N_DEMAND_KG_HA["veg"]=150 — use a crop-specific range, not a single constant; residue-N (broccoli 183 kg N/ha, 71% of uptake) couples to the excreta/residue nutrient loop. ⚠ The ~135 kg N/ha whole-garden proxy (Journeaux 2019) and the AHDB/Feller figures are leads only (primaries not held). CR_056
  • NZ conventional vegetable agronomy (Reid & Morton 2019 + Curran-Cournane & Rush 2021): commercial per-crop marketable yields (harvest fraction varies hugely — broccoli ~18% vs cabbage ~75%) and a sector-wide ~38.5 t/ha gross average across ~45,000 ha (0.2% of NZ; LUC 1–3 land constraint; Pukekohe −46–55% by 2043 to urban sprawl). CONVENTIONAL data — the contrast baseline, NOT the Neobiome veg_yield (regenerative-only, D_001). The system-agnostic takeaways the model uses: the diet-diversity frame (5×75 g/day, MDD-W groups; legumes + dark-green-leafy the binding gap) and the environmental critique of intensive production that justifies the regenerative selection. OT_069 LIT_056
  • Permaculture design system (Mollison’s 12 principles) functions as the primary integrated food production framework in ecovillage contexts — site observation, closed-loop nutrient cycling, food forest design, and polyculture intercropping are the foundational methods for community food self-sufficiency. LIT_012
  • Anaerobic-digestion digestate as biofertiliser (nutrient loop). Where a community runs a community-scale digester on effluent/food waste, the digestate is a nutrient-rich biofertiliser — the stronger justification for AD than its (modest, heat-only) energy yield, and the food-system link that closes the waste→nutrient loop. Dairy/livestock effluent is the typical feedstock. Digestate is ~82–90% plant-available ammonium-N and qualifies as an ACVM-exempt biofertiliser if BANZ-accredited (else a biosolid); a community applying its own on its own land falls under regional-plan rules (RT_271 resolved → CR_041). See anaerobic_digestion and the D03 framing. CR_039
  • Digestate-as-biofertiliser, field evidence (Bozeman MT, Ebel et al. 2025). A 12-household cool-climate case study that ran household-scale AD explicitly for biofertiliser, not biogas — closing the food→waste→nutrient loop at the household scale. Over ~12 weeks the digestate’s N, P, K, S, pH and conductivity all increased significantly (p < 0.05); household food waste balanced across fruit + vegetable fractions gave AD-suitable C:N (veg 12.1–18.6, fruit 20.8–25.7; optimal 20–30). Participants applied the digestate to home gardens and lawns (leafy greens, basal + foliar). Corroborates the digestate-nutrient-recovery justification that makes AD defensible in the food domain. ⚠ Montana nutrient values (mean digestate C 2,093 / N 247 mg l⁻¹) are illustrative, not an NZ calibration. See anaerobic_digestion and the D03 framing. LIT_087
  • Crop-residue / grass feedstock + digestate NPK substitute — the food↔energy↔nutrient link (Malawi, Robin & Ehimen 2024). In this AD techno-economic model the co-digestion feedstocks that flip the economics positive are on-farm crop residues — grass silage and maize residue (66% of the corn crop is stalk/silage waste) — and the digestate is valued as a saleable organic NPK-fertiliser substitute (urea/superphosphate/muriate potash at ~$0.46/kg), i.e. the same waste→nutrient loop this page frames for community digesters. Feedstock characterisation used (literature values, Table 2): CH4 yield 0.46 (cow dung) / 0.07 (grass silage) / 0.35 (maize silage) / 0.32 (human faeces) m³ CH4/kg VS; C/N 19.2 / 22.9 / 31.3 / 10.7. ⚠ Malawi context — the crop-residue-valorisation mechanism transfers, the yields/costs do not. See anaerobic_digestion and the D03 framing. LIT_076
  • Small-livestock production benchmarks (CR_043, resolves RT_277 — the protein-module gap). NZ figures to size a community protein contribution: laying hens ~280–300 eggs/yr each on ~45 kg feed/bird/yr; hill-country grazing ~7.5 SU/ha (4.1 sheep + 3.3 cattle SU; OT_082) producing ~137–147 kg carcass meat/ha/yr → a derived ~20–29 kg edible protein/ha/yr (⚠ protein/ha is derived — no NZ source publishes it; 1 SU ≈ 550 kg DM/yr is a separate convention, not in OT_082). Lets the food model move beyond veg-yield-only. The model differentiates grazing by LUC land class (su_ha[LUC]): the NZLRI/LUC carrying-capacity method (OT_083 — Present Average, sheep-and-beef only, Table 23 bands) calibrated to OT_082’s 7.5 SU/ha at LUC 6–7. Full per-unit regional SU/ha lives only in the printed extended legends (not GIS) → a future precision RT. OT_082 OT_083 CR_043
  • NZLRI South Island LUC Extended Legend — the per-unit/per-region carrying-capacity primary (RD_029, advances RT_281). The South Island volume of the NZLRI extended legends OT_083 names as the only home of the per-unit SU/ha (not a GIS attribute). For each of 219 South Island LUC units it tabulates stock-carrying capacity in SU/ha (ewe-equivalents, annual) at three levels — Present Average / Top Farmer / Attainable Physical Potential — DRY and irrigated, across up to four South Island MAF regions (Nelson incl. Marlborough/Westland · Canterbury · Otago · Southland). This is the region-and-unit precision that would replace the model’s single national su_ha[LUC] approximation for South Island sites (the model uses Present Average). Worked anchors: lowland cropping-plain unit 1c1 = 13 SU/ha (Canterbury, Present-Avg, dry) vs low-capability hill unit 5c1 = 6 SU/ha — a >2× spread within the eastern regions that a flat per-class figure cannot capture. Confirms OT_083’s hard scope: sheep-and-beef pastoral only (not dairy/cropping), annualised, grazing-managed. Design signal: the South Island is ~74% class 6–8 (Figure 1), so ~13% class-1–3 versatile land caps grazing there. ⚠ Partial — North Island legends + the 219-unit extraction into the engine lookup remain (downstream; per-cell OCR re-verify). OT_083 OT_082 RD_029
  • Laying-hen cost cell (④f) — indicative only (CR_044, low): mature hen 30–60, coop 50–200 DIY / 500–3,000 pre-built, run 100–500, running cost ~1–2/bird/month feed (foraging-supplemented) + bedding/health — Dine-A-Chook NZ, WebFetch-verified. **Pasture-grazing cost stays `GAP`** — only the stock-fence rate 33.35/m is sourced (Waikato River Authority schedule, grep-verified); per-ha OPEX/stock/yards/water were unverifiable and dropped (RT_280 stays open). CR_044
  • Pasture-grazing OPEX cell now sourced (④f, OT_091 — resolves the RT_280 gap CR_044 left open). The B+LNZ Sheep & Beef Farm Survey per-ha farm-class P&L (read from the primary): total farm expenditure ~NZD 740–980/ha for hill country (Class 3 hard hill 741, Class 4 hill 977, NNI 2023-24 mean), ~NZD 160/ha extensive high country, ~NZD 1,100/ha intensive finishing. Core recurring inputs: fertiliser NZD 86–144/ha, R&M NZD 65–92/ha (+ animal health, shearing). ⚠ Boundary: total expenditure includes interest (~NZD 167–177/ha), rent and manager wage — a debt-free community on its own land/labour uses a cash-operating sub-total ≈ NZD 575–810/ha. Stocking 7.4–8.8 SU/ha cross-checks su_ha[LUC]. Stock-purchase $/head + stockyards + water-reticulation + a firmer poultry budget stay open (RT_280 partial). OT_091
  • Farm-input price-escalation index (④f, OT_128 — the food analogue to RT_368’s civil-cost escalation). B+LNZ Sheep and Beef On-farm Inflation 2023-24 (P24002) — the annual change in prices paid for farm inputs (NOT a /ha P&L; this is the report [[cr_044_nz-livestock-capex-opex-costs|CR_044]]'s dropped per-ha figures traced to). On-farm inflation **+2.8%** in the year to March 2024 (prior year +16.3%; cumulative **+30.8% / 5 yr, +35.9% / 10 yr**), the single biggest 2023-24 driver **Interest +12.0%** (weight 11.4% → +1.37 of the +2.84 weighted total, about half of all on-farm inflation) — so the headline overstates a debt-free community's escalation; **underlying inflation excl. interest is +1.7%** (+37.0% / 10 yr), the rate to use for a self-sufficient grazing operation. Per-line 2023-24 rates: Fertiliser/Lime/Seeds **−4.2%**, Insurance +8.7%, Animal Health +8.0%, Fuel +4.5%, R&M +2.3%; expenditure weights (Appendix 4) fertiliser 18.0% / R&M 12.2% / interest 11.4% / wages 10.2% cross-check OT_091's per-ha line composition. **Use to de-date the ④f food/livestock cost cells** ([[ot_091_blnz-sheep-beef-farm-survey-2023-24|OT_091]] 2023-24, [[ot_082_beeflamb-hill-country-sheep-beef-2020|OT_082]] 2020, CR_044) to a common base year. ⚠ index only — percentages, no values; NZ-average (N.I. Hill Country Class 4 ran 3.1%). ⚠ Source-internal discrepancy on the underlying rate: Table 2 (p.8) shows +4.4% (index 1,919 vs 1,838) for the 2023-24 underlying figure, contradicting the +1.7% stated in the Summary/narrative/Appendix 3 — +1.7% is retained as the headline rate but the escalation cell built on it is approximate, not exact (see OT_128 Notes). OT_128
  • Auroville (Tamil Nadu): despite available arable land, only 15% of food demand is met internally; 85% externally sourced; agriculture stagnating as economically less competitive than manufacturing and services — land availability alone does not guarantee food self-sufficiency; economic incentive structures are a binding constraint on FSR. LIT_012

Regenerative agriculture (NZ)

  • 11 NZ RA Principles (Lang et al. 2021) applicable to community food system design: maximise photosynthesis year-round (P8), minimise disturbance (P9), harness diversity (P10), manage livestock strategically (P11) — outcomes-based framework compatible with technology integration at community scale. LIT_013
  • RA defined by outcomes, not input restrictions — precision agriculture sensors, soil monitoring, app-based farm data capture, and remote sensing are all compatible with regenerative farming; RA and smart agriculture are complementary, not competing paradigms. LIT_013
  • Remote/proximal sensing and phone apps for farmer-enabled data capture proposed as scalable monitoring pathway — directly relevant to D02 smart food system architecture at community scale. LIT_013
  • Natural capital valuation recommended as the correct economic framework for community food production assessment — standard output metrics understate full return from regenerative food systems by excluding soil carbon, biodiversity, and water quality improvements. LIT_013

European ecovillage food technology adoption (N=38)

  • Dominant food production methods: composting ~90%, land cultivation ~85%, soil regeneration ~65%, food forests ~45%, wild foraging ~45%, animal husbandry ~35%, agroforestry ~25%, aqua/hydroponics ~8%, vertical farms ~0% — low-tech biological methods dominate; no surveyed European ecovillage uses vertical farms; food forests and wild foraging are near-equal in prevalence. LIT_020
  • Regenerative agriculture — definition (Newton et al. 2020): there is no agreed/regulatory definition of regenerative agriculture; across 229 papers + 25 practitioner sites it is defined by processes (cover crops, livestock integration, reduced/zero tillage), outcomes (soil health, carbon sequestration, biodiversity), or both. Implication for a Neobiome food spec: define it explicitly for our own context — a chosen process suite → stated outcomes — rather than relying on the label. LIT_046
  • Ecovillages are “neither primitivist nor modernist” — they use technologies that are available and serve their purposes; the strength of ecovillage food systems lies in socio-technical communal integration, not in technological novelty or sophistication. LIT_020

Multi-scale food production in eco-cell design (Auckland)

  • Eco-cell food integration at three scales: apartment balcony growing space (all bedroom configurations designed with growing balcony), building-scale vertical gardens and rooftop growing, and block-scale local farm programme — food production designed into the urban fabric at every level, targeting 90% local food provision. Auckland feasibility study; proposed figure, not empirical outcome. LIT_018

Vegetable SS benchmarks at neighbourhood scale (Knivsta, Sweden)

  • Vegetable SSI: 38% in dense multifamily (72 m²/person green area) → 210% (same population, more green area) → 400% (low density, 67,000 m², 102 persons). Green area per capita is the binding variable — density and green area share are the dominant design levers simultaneously. LIT_022
  • Land-based production method contributions to vegetable demand (Case 1): biointensive ground farming 25%, standard ground farming 13%, forest gardens 10%. LIT_022
  • Vegetable SSI ≠ full food SS: even 400% vegetable SSI does not cover full dietary requirements — other food categories (animal products, grains, oils) are not addressed by this metric. LIT_022

NZ community food benchmark (Wellington)

  • NZ design-derived benchmark: 200m²/person for plant-based food self-sufficiency using permaculture zoning (5-zone system), raised beds, food forests, and beehives. Not empirically measured — derived from 10-case homestead synthesis and Paremata design iteration. OT_018
  • Beehives essential: fruit trees produce 3× more fruit with beehive than without — direct productivity multiplier for food forest systems. OT_018

Earthsong dense-urban food outcome (INT_006, NZ lived practice)

  • A real NZ outcome against the design-figure benchmarks above: ~180 fruit trees in common land (excluding private gardens), organic + permaculture, doing well for fruit and leafy greens but growing little or no potatoes, root vegetables, or grains — the category gap that vegetable-SSI metrics (LIT_022) systematically miss. Full food self-sufficiency was never the intent on a dense-urban site: “we don’t need to be because we’re part of the wider society” — the same deliberate demand-reduction-not-autonomy stance Earthsong takes on energy and water. Interview VI [INT_006]

Biodynamic practice & community-scale food self-sufficiency (INT_007, practitioner)

  • biodynamic_farming — first practitioner account of biodynamic/regenerative method (closed cow→soil→plant nutrient circle, no synthetic inputs, rotational grazing, certification via Demeter/Skal). The binding constraint vs. industrial farming is labour (handpicking vs. spray-and-finish), partly eased by emerging solar-powered laser/mechanical weeding. Candid caveat: anthroposophic elements (preparations, lunar timing) lack scientific proof. Interview VII [INT_007]
  • Community-scale benchmark (practitioner estimate): ~20 people can be fully food self-sufficient (bar non-local items) on ~3–8 ha (animals, fruit trees, grain for bread); a ~1 ha market garden covers a year’s vegetables for 20 — grounding the design-figure benchmarks (OT_018 200 m²/person; LIT_022 Knivsta) against lived practice. Scale is flexible (backyard → ~1,500 ha). NZ-recalibration pending (RT_196). Interview VII [INT_007]
  • Ideal farm-settlement layout (NI design template): forest integration (~5% cut/yr for wood + heat), compost-pile heat piped to houses, a pollinator flower-ring, fruit-tree/perennial-herb paths, a water-wheel grain mill, and beaver-dam analogues — food, water, energy, and biodiversity designed as one system. Interview VII [INT_007]

NZ food-system model constants (veg demand, garden cost, irrigation)

  • Per-capita vegetable demand: ~47 kg/person/yr actual (2008/09 NZANS) vs ~82 kg/person/yr guideline (MoH 3 servings/day); FAO supply 104–130 incl. waste. Use 82 kg/person/yr as the ”% of veg needs” denominator (low confidence — no NZ nutrition survey since 2008/09). CR_018
  • D29 food_ssi blend-weight basis — NZ macronutrient dietary-energy split (RD_030, resolves RT_302). The 2008/09 NZ Adult Nutrition Survey’s percent-energy shares: protein 16.4% (M) / 16.5% (F), total fat 33.7% / 33.8%, carbohydrate 46.0% / 47.1%, alcohol 3.8% of dietary energy (all within their AMDRs; median energy 10,380 / 7448 kJ). This is the calorie-share basis for the engine’s food_ssi = w_veg·veg + w_protein·protein weights (D29, params.py food_blend_w_veg/food_blend_w_protein, interim 50/50, AF-23): protein supplies ~1/6 of NZ dietary energy → down-weight the protein axis (protein ≈ 0.16–0.26 depending on normalisation, veg = complement), not co-equal. ⚠ The engine axes are food-PRODUCTION axes (veg-food vs protein-food SSI), not macronutrient axes — and vegetables/fruit are minor energy contributors (fruit 5%, potatoes 6%; most energy is bread 11% / grains 7%) — so the split fixes the protein share cleanly but the veg-axis mapping stays a modelling judgment; keep the weight assumed with RD_030 as its anchor. Weights unchanged pending re-derivation (WHO-split corroboration now in-corpus — OT_165, WHO/FAO TRS 916: NZ protein 16.4–16.5% sits within the WHO 10–15% goal). Primary-verified (pdftotext). RD_030
  • WHO/FAO healthy-diet macronutrient REQUIREMENT basis for food_ssi (OT_165, WHO TRS 916 2003). The Joint WHO/FAO Expert Consultation’s population nutrient intake goals (Table 6), as % of total dietary energy: total carbohydrate 55–75%, protein 10–15%, total fat 15–30%, free sugars <10%, saturated fat <10%, trans fat <1%, plus fruits & vegetables ≥400 g/day and salt <5 g/day (sodium <2 g/day) OT_165. This is the authoritative requirement (healthy-diet target) side for the proposed macronutrient food_ssi (carb/protein/fat streams, SSR_macro = min(1, produced ÷ required)) — each macronutrient requirement = (people × daily energy) × (% of energy) ÷ (kJ/g). It corroborates and contextualises RD_030 (closing that bullet’s “WHO-split corroboration running separately” flag): NZ actual protein 16.4–16.5% sits just above the WHO 10–15% goal, NZ carbohydrate 46–47% below the WHO 55–75% goal, NZ fat 33.7–33.8% above the WHO 15–30% goal — so the observed NZ diet is higher-fat/lower-carb than the WHO healthy target, and the requirement basis the model picks (WHO healthy goal vs NZ actual) moves the carb/fat denominators. ⚠ These are POPULATION goals for chronic-disease prevention (a normative healthy diet), NOT an observed split or individual RDIs; carbohydrate’s band is the widest because it is the balancing macronutrient (Table 6 footnote b). Requirement basis only — no engine parameter changed. OT_165
  • food_ssi demand-side REQUIREMENT basis — WHO 2023 macronutrient guideline set (OT_166, international healthy-diet targets). WHO’s current population nutrient intake goals: total fat ≤30% of total energy (supersedes TRS 916’s 15–30%E band), saturated fatty acids ≤10%E (further <10%E conditional), trans-fatty acids ≤1%E (further <1%E conditional), carbohydrate primarily from whole grains, vegetables, fruits and pulses, ≥400 g/day of vegetables + fruits (adults; children 250/350/400 g by age band) ≈ 146 kg/person/yr, and ≥25 g/day naturally occurring dietary fibre (adults; children 15/21/25 g). This is the international aspirational healthy-diet demand anchor for D29 food_ssi = production ÷ demand — sitting normatively ABOVE the NZ-actual basis (RD_030 macronutrient split; CR_018 ~47 kg/yr veg): the WHO ≥400 g/day fruit+veg target is ~3× NZ-actual veg, and NZ actual fat 33.7%E sits just above the WHO ≤30%E ceiling. ⚠ Aspirational vs actual — the WHO set is a demand-basis SENSITIVITY anchor, not the operative demand; and it states no protein %-energy goal, so it does NOT re-weight food_blend_w_protein (that stays RD_030/RT_302). No engine cell auto-changed. Verbatim WHO guidelines, pdftotext-verified. OT_166
  • International macronutrient REQUIREMENT basis — FAO/WHO AMDRs (FNP 91, the requirement counterpart to RD_030’s NZ-actual intakes). The Joint FAO/WHO Expert Consultation (FAO Food and Nutrition Paper 91, 2010) sets the population acceptable macronutrient distribution ranges the food_ssi demand side is measured against: adult total fat 20–35%E (minimum 15%E, or 20%E for women of reproductive age / underweight adults), SFA ≤10%E, total PUFA 6–11%E, n-6 LA 2.5–9%E, total n-3 0.5–2%E with an absolute EPA+DHA 0.250 g/day, TFA <1%E (children 2–18 yr total fat 25–35%E, SFA 8%E). RD_030 confirms NZ adults actually eat fat at 33.7/33.8%E — near this range’s upper bound. ⚠ Two design flags, not new cells: (1) the model’s veg+protein production axes do not represent dietary fat/oil adequacy, yet FNP 91 requires ≥15–20%E from fat — a fat-deficient production mix would not be flagged by current food_ssi; (2) the EPA+DHA 0.250 g/d requirement is seafood-met, a genuine adequacy gap for inland land-based communities. ⚠ Scope: FNP 91 is fats/fatty-acids ONLY — it carries no protein or carbohydrate requirement (companions WHO TRS 935 / TRS 916). Authoritative FAO/WHO primary, verbatim (pdftotext). OT_167
  • Organic community market garden: establishment ~100–150k/ha; **operating ~90–130k/ha/yr, labour-dominated** (1.5–2 FTE; commercial ~14 FTE/ha) — operating dwarfs capex, so food-domain economics need an explicit operating/labour term, not a capex %. Irrigation ~4.5 ML/ha/yr is the CONVENTIONAL figure — contrast only under D_001, NOT used for the coupling (superseded by CR_033’s regenerative ~2.0 ML/ha/yr). CR_018
  • Grounds INT_007’s “~1 ha feeds 20 people” benchmark: 82 kg × 20 ≈ 1,640 kg/yr veg need. CR_018
  • Food↔water coupling — regenerative irrigation (CR_033, resolves RT_259): a regenerative/permaculture NZ veg garden needs ~2.0 ML/ha/yr applied irrigation (200 mm ≈ 45% of conventional 4.5) — point estimate, low–med confidence, no NZ peer-reviewed figure (built-up). Region-dominated: wet-west <1.0, dry-east 2.5–3.0 (summer PED 300–500 mm) → the model should scale by regional PED, not apply 2.0 flat. Mulch cuts the evaporation slice (~15–30% off seasonal irrigation, not the popular 70%); GROW BIOINTENSIVE’s “88% less” is per-pound advocacy, not area-based. Under D_001 the conventional 4.5 (CR_018) is not used. CR_033
  • Plant-protein cell (④f, D29) — NZ faba bean primary (OT_088, resolves RT_289). FAR Focus 8 Faba beans: A growers’ guide (Dec 2012) — a 55-paddock Canterbury/North-Otago survey: mean grain yield 5.6 t/ha (range 1–8.5; NZ crops 1–9 t/ha) at 27.7% (spring) / 31.1% (winter) whole-seed crude protein → ~1,550–1,740 kg edible protein/ha/yr (single annual crop). Lifts the food model’s interim LEGUME_PROTEIN_KG_HA 900 → ~1,500 kg/ha (conservative spring end; absorbs moisture-basis uncertainty). ⚠ FABA-specific conventional arable (not a measured biointensive plot); a mixed pulse plot (peas ~0.7 t protein/ha) sits lower → faba is the upper-realistic NZ value, RT_291 raised for a field-pea/lupin primary. Whole-seed crude protein, not digestibility-adjusted. Newer NZ sources corroborate not supersede (2006 top cultivar 5.7 t/ha; Navneet 2025 varieties ~24–27% protein). OT_088
  • Pulse-protein spread placing faba as the upper-realistic NZ anchor (CR_054, medium — corroborates OT_088, no new cell). The RT_289 synthesis behind OT_088 situates faba (~1,550–1,740 kg protein/ha, the model’s LEGUME_PROTEIN_KG_HA anchor) at the top of the NZ pulse spread: field peas ~3.5–5.0 t/ha × 18–23% → ~0.7–1.0 t protein/ha (NZ national implied ≈0.7, matarau.nz), lupins 22–41% DM protein but lower NZ grain yield → below faba. Confirms the constant stays conservative (faba = upper-realistic, not a mixed-plot floor) — consistent with OT_112’s chickpea/buckwheat low-enders. ⚠ whole-seed crude protein, not digestibility-adjusted; field-pea/lupin primary still open = RT_291. CR_054
  • NZ faba protein cross-check (Navneet et al. 2025 — independent NZ measurement behind the plant-protein cell). Four NZ-grown faba varieties (Early Long Pod, Evergreen, Coles Dwarf, Janet), Canterbury 2022–23: flour crude protein 24.81–27.49% (N×6.25), starch 35.3–39.1%, dietary fibre 9.7–10.3%. Sits at/just below the lower end of OT_088’s whole-seed 27.7% (spring) / 31.1% (winter), corroborating the conservative ~1,500 kg protein/ha/yr the model takes from OT_088 (Navneet is sieved flour, OT_088 whole-seed — a bracket, not a replacement). Reports no yield (t/ha) — firms only the protein-% multiplier of LEGUME_PROTEIN_KG_HA, not the yield term. Food→energy aside: overnight soaking cut cooking time “by at least 30 min” / “nearly a third” across all four varieties, “thereby decreasing energy consumption” (variety-dependent: Janet ~73 min raw vs Evergreen ~57 min). LIT_088
  • NZ commercial vegetable-crop yields (CR_051 — food-domain validation anchor, RT_307). Published NZ govt/industry yields: potatoes 45 t/ha (MPI GHG inventory long-run avg) / 51 t/ha (Potatoes NZ commercial 2022) / 64 t/ha (high-input, Tupu/MPI); onions 45 t/ha (Ecoinvent/MPI 2009–12). These bulk-staple monoculture yields sit far above the mixed biointensive-garden figure (~21–25 t/ha, LIT_058 / CR_021) — the expected staple-vs-garden gap. ⚠ Confirms no NZ permaculture/CSA per-ha yield dataset exists (RT_307 gap); commercial crop stats are the closest proxy. Primary → RT_323. AI-compiled → medium. CR_051 — potato 45 t/ha (GHG-inventory) / 48 t/ha (industry) now primary-verified via OT_096 (MPI Technical Paper 2025/05); Potatoes NZ 51 / Tupu 64 / onion 45 stay secondary. OT_096
  • Tupu/MPI potato fact sheet — the industry primary behind CR_051’s 64 t/ha (OT_108, resolves the Tupu leg of RT_323). The Te Puni Kōkiri / Tupu.nz Land use fact sheet: Potatoes states the current national commercial average yield “now sits at an average of 64 tonne per hectare” (framed as 20-year growth to present, not a high-input outlier), giving gross revenue 25,600–44,800/ha at $400–700/t — primary-verified verbatim (verified). Sits well above the mixed biointensive-garden veg_yield (~21–25 t/ha, LIT_058 / CR_021) — the expected staple-vs-garden gap, not a contradiction. Reconciles with OT_096’s constant 45/48 t/ha by vintage/purpose (fixed GHG-inventory constant vs a current industry average): cite 45–48 for a conservative basis, 64 for a current-commercial upper anchor. OT_108
  • Commercial-potato establishment economics + agronomy (OT_108, ④/cost + land-class layers). **5,000/ha to establish and grow** (spray/fertilise/cultivate 2,200 + seed potatoes 2,500 + contractor-plant 150; harvest is ADDITIONAL, excluded) — a commercial-monoculture contrast/upper baseline against the labour-dominated organic community market-garden operating cost ($90–130k/ha/yr, CR_018), a wholly different (hand-scale, high-diversity) production mode. Agronomy corroborates the model’s arable gating: LUC 1–4, deep well-drained soils, 500–700 mm water over a 120–150 day season, 18–20 °C optimum (10–30 °C tolerance, plant after frost). Bulk-staple validation anchor, not a model input. OT_108
  • Legume/pulse land-class rule — RT_315 SETTLED (CR_052). The LUC system draws a hard line: LUC 5–7 are not suitable for arable cropping (LUC Handbook 3rd ed, Landcare); harvestable pulse/grain crops (soy, chickpea, lentil, lupin, faba — the model’s legume_protein_kg_ha crop, OT_088) need flat/well-drained/machinery-accessible land → LUC 1–3, or explicitly screened LUC 4 (slope <12°, free-draining). A ≤1 ha pulse-grain patch on LUC 5–6 cannot yield harvestable grain “under any management” — the limit is physical, not managerial. Forage legumes (clover, lotus, lucerne; oversown, no cultivation) are a separate pathway — viable on LUC 4–6 pastoral for N-fixation + feed (15–25%+ DM, Hill Country Futures), NOT modelled as harvestable protein. Confirms the engine’s existing arable cap — do NOT extend the veg-garden LUC exemption (D29) to harvestable pulse grain; the honest protein gap on all-pastoral parcels is correct. The ‘LUC 5–7 not suitable for arable cropping’ statement is now primary-verified via OT_097 (LUC Handbook 3rd ed, Lynn et al. 2009). CR_052 OT_097
  • NZ specialty grain/pulse land-class set — six viable crops (OT_112, resolves RT_326). LFI / Our Land & Water NGFS Specialty Grains and Pulses Report (Jun 2019) shortlists oats, quinoa, chickpeas, soybeans, buckwheat, hemp as crops that “can be grown in current pulse and grain growing regions of NZ”, with compiled Table 4 yields — oats 5–8, quinoa 3–5, soybean 2–5, chickpea 2–4, buckwheat 1.5–2 t/ha; hemp 0.8–1.0 t/ha seed — and whole-seed protein for two (chickpea 20%, buckwheat 13–15%). Yield × protein bounds the plant-protein constant with a crop set: chickpea ≈ 400–800, buckwheat ≈ 200–300 kg protein/ha/yr — all well below the faba anchor’s ~1,550 kg/ha (OT_088) → confirms LEGUME_PROTEIN_KG_HA stays conservative (faba = upper-realistic, not a mixed-plot floor). ⚠ COMPILED agronomic ranges (seed-company / FAR / overseas), not new NZ trials. OT_112
  • All six confined to NZ’s established arable regions — reinforces the LUC 1–4 arable gate (OT_112). Table 4 places every shortlisted crop in Southern North Island, Canterbury, Southland, South Island (good arable land, broadly LUC 1–4), not marginal grazing land — an independent corroboration of the RT_315 arable-cap verdict (CR_052) that harvestable pulse/grain needs flat, machinery-accessible land. Inputs split the set: soybean + chickpea are N-fixing (regenerative fit, D_001) vs hemp 100–130 kg N/ha; on water, buckwheat “very drought sensitive” / soybean “needs irrigation” vs hemp low-water 250–300 mm — feeding the same regional-PED irrigation logic as veg (CR_033). OT_112
  • Measured NZ multi-site hemp trial firms OT_112’s compiled hemp seed range — on the OIL path, not protein (OT_168, McPartland/Cutler/McIntosh 2003). NZ’s first two licensed hemp seasons — 11 sites/55 ha (2001-02) → 19 sites (2002-03), 11 cultivars — measured seed yield averaging 950–1,800 kg/ha (max 2,800) and fibre stalk biomass (dry matter) 6,000–11,000 kg/ha (max 13,900) OT_168. The measured seed low end (~950 kg/ha) sits right at OT_112’s compiled hemp high end (0.8–1.0 t/ha) and extends well above it — so measured NZ trial data firms and modestly raises the hemp seed-yield figure (the measured counterpart to OT_112’s seed-company/FAR/overseas range), the role OT_088 plays for faba. ⚠ Seed protein is NOT measured — hemp seed’s food product here is cold-pressed oil; protein is qualitative only (“high in protein and omega-3 fatty acids”), profiling deferred to future work — so this supplies no hemp protein constant and does not disturb the faba plant-protein anchor. Independently restates the LUC 1–3 arable gate (“annual crops such as hemp can only be sustained on LUC Class I, II, or III soils”, 3.8 M ha/14% of NZ) and a 82–150 day sowing-to-harvest window (cross-checks OT_112 ~100–110 grain days; frost-free/GDD-gatable via RD_022). ⚠ 2003 vintage; N-uptake figure cites Hall 2000 not the NZ trials; fibre DM is a biomaterials/biomass figure, not food. Measured NZ multi-site trial, verbatim (pdftotext). OT_168
  • Hemp seed yield corroborator + land-use economics (OT_169, Perrin Ag “Potentially Hemp”, NZAGRC 2023 — COMPILED, medium). The fourth NZAGRC alternative-land-use report independently reports NZ hemp seed yield 0–2,000 kg/ha, average ~800–1,000 kg/ha (cited to Tupu.nz 2022 + Marsh 2020) — a second compiled witness to OT_112’s hemp 0.8–1.0 t/ha seed, and the low end of the measured NZ envelope now in-corpus (NZGA Fasamo 804–1,849 kg/ha; NZHIA 950–1,800, max 2,800) → confirms 0.8 t/ha as a defensible conservative anchor, not a contradiction. Land-use economics (④f context, NOT a wired cell): seed gross margin ~2,336/ha**, operating surplus **1,725/ha, but 50 ha seed needs **1,265,000** capital → **IRR 4.1%** (fibre 877,500 → 6.7%; both far below milling wheat 12.8% and NZ dairy EBIT $3,189/ha) — hemp is capital-intensity-limited, not agronomically-limited, and uncompetitive with dairy/wheat on the same LUC 1–4 land. ⚠ OIL-not-protein: this source measures nothing on hempseed protein — Table 1’s 30 g/100 g is a USDA reference, NOT an NZ measurement; hemp is an oilseed (1 t/ha seed → ~250 L oil + ~750 kg meal) needing 100–130 kg N/ha, so it is not the model’s plant-protein anchor (that stays faba, OT_088). NZ-measured hempseed protein % remains unpublished. OT_169
  • MEASURED NZ hempseed yield — firms OT_112’s compiled hemp range (LIT_089, Townshend & Boleyn 2010). A two-season Canterbury field trial of industrial hemp cv. Fasamo (Midlands Seed Ltd, ASNZ Special Pub. 13 / Grassland R&P Series 14): machine-dressed seed yield (8% MC) 804–931 kg/ha on a marginal border-dyke/effluent site (2006-07, Ashburton) and 1,640–1,849 kg/ha on a well-prepared arable site (2007-08, Methven) — the between-site/season difference dwarfing plant population (125–250 plants/m²; recommended 150–225). This measures what OT_112 gives only as a compiled range (800–1,000 kg/ha seed): OT_112’s figure matches the marginal-site low end, a good Canterbury arable site roughly doubles it → cite the measured NZ range ~0.8–1.0 (conservative) to ~1.6–1.85 t/ha (good arable), not a single point. The hemp analogue of what OT_088 did for faba. ⚠ Quality axis measured = OIL, not seed protein: measured seed oil content 30.9–32.5 g/100g + fatty-acid composition (linoleic ~53%, α-linolenic ~20%, oleic ~10%, γ-linolenic 3.8%; AOAC 991.36) approaching the ~3:1 linoleic:α-linolenic EFA balance — firms the dietary-fat/oil-adequacy axis (OT_167 15–20%E fat floor), NOT the plant-protein cell; no hempseed protein % is measured here. Both trials on irrigated Canterbury arable (LUC 1–4) + fertilised (~100–130 kg N-equiv) → corroborates OT_112’s arable-region + higher-N placement (poorer regenerative fit than N-fixing pulses, D_001). Measured NZ primary, verbatim (pdftotext). LIT_089
  • NZ plant-protein land-suitability criteria corroborate the arable gate + add explicit climate thresholds (OT_164, PFR/MPI 2018). Plant & Food Research’s Opportunities in plant based foods – PROTEIN (SPTS 15748, for MPI) applies a GIS scan for high-value plant-protein crops using LUC 1–3, slope <5°, growing-degree-days ≥800 (base 10 °C), and frost-free period >180 days, estimating >1,737,000 ha of NZ land suitable (>10× the ~140,000 ha in horticulture). This is an independent NZ confirmation of the model’s arable cap (LUC 1–3 highly-productive pool, AF-41; CR_052, OT_097) and adds two thresholds the model does not yet encode explicitly — a GDD-800 floor and a frost-free-180-day floor (implementable from the already-ingested NIWA climate layer RD_022). The slope <5° is a high-value-land tightening of the general <12° arable screen, not a contradiction. It also widens the NZ-growable protein-crop set (adds alfalfa/lucerne — leaf protein ‘up to 30% by weight’, a leaf-protein-concentrate candidate distinct from the whole-seed pulse anchors, OT_088) and the by-product protein streams (spent grain/wheat bran 19–30% w/w) behind its ‘whole-of-plant / total utilisation’ economics. ⚠ Advocacy/opportunities report; the 1,737,000 ha figure is the report’s own ‘rough estimate’ GIS output, not a validated dataset; alfalfa is leaf (and a NZ forage), a candidate flag, not a model input. OT_164

🔴 The arable pool this domain optimises for is the land planning law protects FROM dwellings (REG_009)

The NPS-HPL defines highly productive land as LUC 1, 2, 3 — the same arable pool the food model rewards above (arable_pct_luc13). And it directs councils to avoid it being built on:

  • cl 3.9(1): “Territorial authorities must avoid the inappropriate use or development of highly productive land that is not land-based primary production” — and “land-based primary production” is defined as production “reliant on the soil resource of the land”. Dwellings are not.
  • cl 3.8(1): “Territorial authorities must avoid the subdivision of highly productive land” (unless the lots retain productive capacity, or it is specified Māori land, or specified infrastructure).
  • Policy 6: rezoning and development of HPL as rural lifestyle is avoided.

So a site can pass the food gate and fail the planning gate — and the engine cannot currently see it (RT_344). The sharpest framing: the law says grow food there, don’t live there. Note cl 3.9(2)(aa) expressly permits “intensive indoor primary production or greenhouse activities” — the community’s food infrastructure is fine; it is the houses that are constrained.

⚠ Do not encode this as “LUC 1–3 = agriculture only”. The NPS-HPL says avoid, not prohibit: exceptions survive for specified Māori land, “a small-scale or temporary land-use activity that has no impact on the productive capacity of the land”, “supporting activities”, and s6 matters. The honest treatment is a planning-risk / consent-difficulty flag, not a feasibility kill.

🟢 The Māori-land pathway escapes both — specified Māori land is an express exception to cl 3.8 and cl 3.9, and the NES-Papakāinga separately permits ≤10 homes on ancestral Māori land. It is the only doubly-privileged route to a community on good soil. REG_009

Community dairy — the NZ edible-fat pathway

  • Per-cow feed DEMAND, and why the printed number is not the planning number (OT_171, DairyNZ Facts & Figures ch.4, Dec 2021). Annual dry-matter requirement at 11.0 MJ ME/kg DM: Jersey 375 kg LW @ 300 kg MS = 3.8 tDM/cow/yr · Kiwicross 450 @ 400 = 4.8 · Friesian 500 @ 400 = 5.1, including walking 4 km/day for 270 days in milk OT_171. 🔴 These are EATEN figures, not GROWN — verbatim “They do not allow for any feed offered that was not eaten (wastage)” — and the source’s own worked example uses 80% utilisation, so 5.1 tDM eaten ≈ 6.4 tDM grown; a further +5% per MJ ME below 11.0 applies to low-input pasture, which by definition sits below the base. Net: taken raw the 3.8–5.1 band understates a low-input community’s real pasture need by roughly 25% — treat it as a floor, not a planning figure. ⚠ The chapter’s own mature liveweights (Jersey 435 kg / Friesian 535 kg) are ABOVE the 375/500 kg rows the upstream research report selected, a second understatement. Converting feed demand to a land area still needs a regional pasture-growth (t DM/ha/yr) figure the corpus does not hold (RT_389). OT_171
  • The fat cell — kg milkfat per cow, published directly (OT_172, DairyNZ/LIC NZ Dairy Statistics 2023-24). National basis 10,485 herds · 4,701,596 cows · 1,703,404 effective ha · 2.76 cows/ha · 400 kg MS/cow (225 kg milkfat + 176 kg protein) · 1,105 kg MS/ha; Table 4.6 herd-test by breed: Jersey 3,669 L/cow, 208.7 kg milkfat, 5.71% fat · HF/J cross 4,765 L, 238.3 kg, 5.05% · Holstein-Friesian 5,250 L, 236.4 kg, 4.56% · Ayrshire 4,766 L, 209.1 kg, 4.42%; Table 4.7 liveweight Jersey 414 / HF-J 492 / HF 537 kg; breed share HF/J 60.4%, HF 23.9%, Jersey 7.5% OT_172. 🎯 Because milkfat is published in kilograms per cow, the edible-fat model needs no milk-density assumption and no g/L conversion. Liveweights key straight into OT_171’s feed table — note Jersey 414 kg sits between OT_171’s 400 and 425 kg rows, not at the 375 kg row the upstream report used. ⚠ The 2.76 cows/ha is a methodology artefact, not destocking — the publication’s own note reports the 2023/24 collection method changed, giving “larger-than-usual variations in total effective hectares, average farm size and stocking rate”; the prior decade ran 2.81–2.87, so use ~2.8 as the commercial benchmark. ⚠ Commercial twice-a-day figures = a ceiling for a community system (once-a-day counterpart: LIT_090). ⚠ Milkfat ≠ milksolids — only the 225 kg/cow milkfat component is available as edible fat; reading MS as fat overstates fat ~80%. Dairy cost leg still absent (RT_387); milkfat→butter recovery still unsourced (OT_173, RT_388). OT_172
  • Milkfat → butter conversion, and the term that is still empty (OT_173, FAO APHP 69 Village milk processing, 1988). Butter is costed at 82% fat — verbatim “the cost price of 1 kg of butter will be based on the cost of 820 g of fat content” — giving butter_kg = milkfat_kg × recovery ÷ 0.82; village cheese yield “approximately 11 kg of cheese per 100 litres of milk” (Gouda-type) OT_173. 🔴 Two corrections travel with this source. (1) Scope: it is a village dairy cost-accounting manual, NOT a compositional standard — the 82% sits inside a worked costing example; the proper instrument is Codex STAN 279-1971 or the NZ Food Standards Code. (2) The 90% fat-recovery figure the upstream research report applied is NOT in this source and is uncited anywhere — so the recovery term of the chain cows × kg milkfat/cow ([[ot_172_dairynz-dairy-stats-2023-24|OT_172]]) × recovery ÷ 0.82 remains unsourced (RT_388). Until it is filled, report community butter output as a band across a stated recovery assumption, never as a point estimate. ⚠ 1988, village-scale, non-NZ. OT_173
  • A house cow is a ONCE-A-DAY cow — the correction that turns commercial dairy stats into a community figure (LIT_090, Lopez-Villalobos et al. 2023, Massey + LIC). Full-lactation once-a-day (OAD) vs twice-a-day: −27% milk, −23% fat, −24% protein, at higher fat and protein concentrations and with better reproductive performance; NZ measured absolute yields (Table 1, kg not litres): Jersey OAD 2,211–2,592 · HF OAD 2,879–2,914 against TAD Jersey 2,839–2,929 and TAD HF 3,824–4,234; Jersey is the most OAD-tolerant breed (−17% vs HF −19 to −25%); OAD is practised whole-lactation in ~10% of NZ herds LIT_090. 🎯 The fat penalty is smaller than the volume penalty — a community optimising for butter loses less than one optimising for milk volume, and Jersey’s OAD tolerance compounds with its highest fat % (5.71%, OT_172). Use the measured OAD rows directly; do not discount the TAD rows by hand. ⚠ Even OAD is an upper bound on a true low-input house cow with a calf at foot — no NZ source measures one. 🔴 Supersedes the lifestyle-blog house-cow figure the upstream research report used: its “Jersey house cow 2,835 L” is a twice-a-day commercial number mislabelled as a house-cow yield (blog not ingested). LIT_090
  • Small-block stocking sanity band — and the arithmetic that says use the low end (URL_032, Northland Regional Council). Lifestyle-block guidance: adult cattle 1–3/ha (steep/difficult, rough pasture) to 2–5/ha (easy/flat, good pasture); sheep 8–12 / 13–18; young cattle 2–3 / 3–5; council advice verbatim “Many lifestyle blocks are overstocked… in general it is better to be understocked than overstocked” URL_032. The only small-block stocking guidance in corpus (commercial benchmarks are ~2.8 cows/ha dairy, OT_172, and ~7.5 SU/ha hill country, OT_082). ⚠ Self-described “a very rough guide”, undated, no underlying dataset, Northland-specific (warm, high rainfall, long growing season) — do not apply unmodified to Southland or Central Otago. 🔴 Its own arithmetic breaks at the top of its range: 2 cows/ha implies ~10 t DM/ha/yr pasture growth (plausible), 5 cows/ha implies ~25 t DM/ha/yr — above almost all unirrigated NZ pasture (derived at verification, NOT printed in the source) → the top of the range needs imported feed or irrigation, so use the low end. Exposes the model’s real missing term: regional pasture growth in t DM/ha/yr (RT_389), which would convert OT_171’s per-cow feed demand into hectares directly and retire the need for a rough stocking guide. Cross-check, not an input. URL_032
  • Regional pasture-growth term now sourced — RT_389 resolved (RD_033, DairyNZ Facts & Figures ch.3, 2nd ed Dec 2021). The 49-district-site dataset the two bullets above flagged as the missing term: NZ regional pasture growth 7.6–21.7 t DM/ha/yr, with a low-input planning band ~8–12 t DM/ha/yr (all sites N-fertilised, some irrigated, so treat as an upper reference for an unfertilised community block). This is the term that converts OT_171’s per-cow feed demand into hectares and cross-checks URL_032’s stocking guide (~2 cows/ha ≈ ~10 t DM/ha plausible; ~5 cows/ha ≈ ~25 t DM/ha, above all unirrigated NZ sites). RD_033

RCS operating rules — the raw-milk Notice, now retrieved (REG_032, resolves RT_390)

  • The operational detail REG_031 lacked is now in hand: MPI Animal Products Notice: Raw Milk for Sale to Consumers (9 Mar 2022), the instrument that supplements the 2015 Regulations. REG_032
  • 🔴 Correction that changes the cost picture: the “heat to 70°C for one minute” figure is a mandated consumer LABEL WARNING (cl 8.4), NOT a producer heat-treatment obligation. The scheme permits raw, unpasteurised sale (label reads “raw (unpasteurised) milk”, cl 8.1). On-farm pasteurisation is optional, not required — so the raw pathway carries a testing + cold-chain + records OpEx burden, not a pasteurisation CapEx line.
  • The recurring compliant-raw burden: ≤6°C cold chain (initial cooling → delivery), 30-hour sell-by (cl 6.9), use-by ≤4 days after milking commenced (cl 8.2), pathogen testing every 10 days for 5 pathogens (Salmonella, Listeria, Campylobacter, Coag-positive Staph, E. coli), reducible to monthly on a clean record (cl 6.11 / Table 6.11), record retention ≥4 years (cl 1.2), plus performance-based verification audits (Part 7). REG_032

🔴 The community-dairy pathway's binding gate is REGULATORY, not agronomic — and it rules out the obvious design (REG_031)

Raw Milk for Sale to Consumers Regulations 2015 (LI 2015/309, in force 1 Mar 2016, version as at 28 Oct 2021) — a dual instrument (reg 3): a regulated control scheme under the Animal Products Act 1999 plus sale/advertising restrictions under the Food Act 2014 REG_031. The binding chain:

  • reg 61 — only a registered farm dairy operator may sell. reg 62 — sale must be direct to the final consumer.
  • reg 64 — hand-over only “at the farm dairy address at which the milk was extracted” or by delivery “to the final consumer’s residential address.” 🔴 A shared village collection point / common-house fridge is NOT lawful unless separately registered as a depot.
  • reg 66 — within 30 hours of the start of milking, held at ≤6°C.

The one structurally compliant configuration is an on-site registered community dairy supplying resident members at the dairy itself. Pasteurising on site removes the pathway from this instrument entirely — a real design fork with its own capital and energy cost, not a preference.

⚠ “Sell” is not defined in these regulations. Whether a community distributing milk to members for a levy or subscription is a “sale” is an open legal question — flag it wherever the pathway is scored; do not resolve it by inference. ✅ The “4-day use-by” and “70°C for 1 minute” details (mis-attributed to these regulations) are now located in the MPI Animal Products Notice, retrieved as REG_032 (RT_390 resolved): use-by = cl 8.2; the 70°C figure is a consumer label warning (cl 8.4), not a producer heat-treatment rule — the scheme permits raw, unpasteurised sale. REG_031

Edible oil crops — NZ pathways

  • Oilseed rape — the ONLY NZ oil crop with both legs from one measured trial (LIT_095, Fasi et al. 2012, Lincoln, Canterbury). Verbatim: “Oilseed rape yielded an average of 2.4 t seed/ha with 42% oil content”; brown mustard and camelina just over 1 t/ha at 30% and 35% LIT_095. 2.4 t/ha × 42% ≈ 1,008 kg oil/ha — IN THE SEED. ⚠ The 2.4 is a mean across four sowing dates (recorded per-sowing values 2,325 / 2,406 / 2,939 kg/ha — sowing date alone moves the answer ~25%) and the trial was irrigated (Wakanui silt loam). 🔴 Oil in the seed is not oil recovered. The trial’s press was “a cylinder hole screen press operating at a throughput of 20-35 g/minute with a choke size of 4 mm and operating temperature of 80°C” — a laboratory analytical press measuring oil CONTENT, from which no recovery fraction can be derived. Do NOT adopt the upstream research report’s “806 kg recovered oil/ha” — it rests on an unverified 80% recovery that no retrieved source supports (RT_391); report oil-in-seed with an explicit flagged recovery band instead. Processing/market legs come from OT_183. LIT_095
  • Olives — productive per TREE, unmodellable per HECTARE (OT_180, Olives NZ 2021 Grove Census). National 10.93 kg fruit/tree at 13.84% oil yield ≈ 1.5 kg oil/tree; regional kg/tree Hawke’s Bay 16.65 · Nelson 12.48 · Wairarapa 12.13 · Canterbury 10.08 · Kapiti 9.52 · Auckland 6.06 · Waiheke 5.50 · Marlborough 4.93 · Northland 3.87 · Central Otago 3.77 — a 4.4× regional spread, so if olives are ever modelled they must be modelled regionally OT_180. 🔴 NO PLANTING DENSITY IS PUBLISHED — no grove area, no spacing, no trees/ha; the word “hectare” does not appear in the document (whole-document search). kg/tree therefore cannot be converted to kg/ha (RT_392), and olives stay a kg-oil-per-tree pathway only. ⚠ Sample 62% response (101 of 163 groves) and verbatim “None of the super groves responded” — ~94,000 trees, almost 30% of the industry, absent; 17% reported “No Harvest” → treat 2021 as the optimistic end of a range, not a central estimate. OT_180
  • Olive series, second point — oil % is a parameter, kg/tree is a range (OT_181, Olives NZ 2024 Grove Census; contested: true vs OT_180). National 5.1 kg fruit/tree (“a significant decrease from 9.2kg per tree in 2023”) at 14.8% oil (excl. Bay of Plenty/Waikato); ~310,000 trees; 343 t harvested, “the lowest total since 2015” OT_181. 🎯 The model-usable finding: kg/tree swings >2× (10.93 / 9.2 / 5.1) while oil yield % stays 13.4–14.8% across five seasons → oil % is the stable parameter; kg/tree must be a range, never a midpoint. Olive oil per tree is therefore ~1.5 kg (2021) to ~0.75 kg (2024) — and per tree only, since neither census publishes trees/ha (RT_392). 🔴 Do NOT read the 2021→2024 fall as a decline: it is season and respondent base and method at once — verbatim “This report includes all trees recorded by census respondents, regardless of whether they were harvested” (2021 declares no equivalent basis), response 45% of 141 vs 101 respondents, “No Harvest” 28% vs 17% — and the national tree count is flat at ~310,000. ⚠ Even the regional ranking inverts (Hawke’s Bay 16.65→5.74; Central Otago 3.77→6.27), so regional point estimates are unsafe without multi-year means. OT_181
  • Hazelnuts — per-hectare yields exist, but a UNIT blocks the oil calculation (OT_182, Hazelnut Growers Association NZ Bulletin 1). Verbatim: “Growers should plan to harvest 1 tonne/ha by year 6 and 2.5 tonnes/ha by year 10”, after a “small crop in the third year after planting, with the first commercial harvest in the fourth or fifth year”; establishment 8,500/ha** (established shelter, no irrigation) to **25,000/ha (bare land); ~400 ha nationally (2007 census), Canterbury then Otago, Nelson, Southland, Marlborough OT_182. 🔴 Do NOT multiply 2.5 t/ha by the 54.6–63.2% oil content (LIT_096): the yields never declare in-shell vs kernel while the oil % is explicitly kernel-basis — the product could overstate oil output ~2× (RT_393). ⚠ The Appleby trial (Nelson 1970-75, 10.3 kg/tree/yr, “over 4 tonnes per hectare based on the spacing used in that trial”) is not transferable — the spacing is never stated. ✅ Usable now regardless of the unit block: the climate envelope — mean annual 12–16°C, minimum −8 to −10°C, chilling 600–1200 h, rainfall 800–1000 mm, with Nov/Dec frosts of −2 to −3°C damaging nut clusters — is directly gateable against RD_022; and the decade-scale time-to-bearing is a real constraint against the annual oilseed-rape pathway (LIT_095). 🔴 Mis-attribution caught: the upstream report’s −13°C dormant / −7°C open catkins are NOT in this bulletin — they belong to the NZ Tree Crops Association hazel guide. ⚠ Its gross margins ($4,000–7,000/ha) are Australia/Oregon/Italy, not NZ. OT_182
  • NZ-grown hazelnut composition — the correct number that cannot be used (LIT_096, Savage & McNeil 1998, Lincoln University). Verbatim “The total oil content of the hazelnuts ranged from 54.6 to 63.2%”, protein 14.3–18.2%, dietary fibre 9.8–13.2%; six cultivars, 12-year-old trees, autumn 1995 harvest LIT_096. On a per-kg-of-kernel basis nuts are far denser oil sources than oilseeds (cf. oilseed rape 42%, LIT_095). 🔴 KERNEL BASIS — established by the paper’s phrase “the remaining portion of the kernel” — which is exactly why it cannot be multiplied by OT_182’s undeclared-basis t/ha yields (RT_393). ⚠ Abstract only (publisher paywalled). Protein/fibre recorded for completeness only — the plant-protein anchor stays faba (OT_088). LIT_096
  • Walnut — the highest oil content in corpus, attached to no yield at all (LIT_097, Zwarts et al. 1999, Lincoln University). Verbatim “Total oil was extracted using a cold press… The total oil content of the nuts ranged from 62.4 to 68.7%”; ten cultivars, replicated trial, 1994 and 1995 harvests LIT_097. 🔴 No yield data at all — and no NZ walnut yield exists at any scale in corpus, so walnut cannot reach a kg-oil figure per hectare or even per tree. A different failure mode from olive (no planting density, RT_392) and hazelnut (undeclared yield basis, RT_393): walnut is blocked on the yield leg itself, which is why no target is opened — an NZ walnut yield is not a document known to exist. ⚠ Abstract only. Oil density is not the constraint on any NZ nut pathway; land-productivity data is — walnut kernel 62.4–68.7% > hazelnut kernel 54.6–63.2% > oilseed rape 42% > olive fruit 13.4–14.8%. LIT_097
  • Screw-press residual oil — the only retrieved evidence on the recovery gap, and it does NOT close it (LIT_098, Burton et al. 2022, non-NZ). Verbatim: “After expelling the oil using a screw press, the oil content in the residue obtained has 8.4–15.5% oil… while the oil content in the de-lipidated hemp meal after solvent extraction is expected to be much lower (~1% oil)”; and the terminology “hemp seed cake should refer to product obtained after mechanical pressing… hemp seed meal is obtained after solvent extraction” LIT_098. What it establishes: mechanical pressing leaves a materially oil-rich cake, so any true recovery fraction is meaningfully below 100%, and the cake is a feed by-product, not waste. 🔴 What it does NOT establish: it is not a recovery fraction (residual oil in cake ≠ fraction of seed oil recovered — no conversion has been performed here), it is hemp-specific (not transferable to rape/olive/hazelnut/walnut presses), and it is non-NZ (Australian authors; Canadian/Italian/Finnish data — topics deliberately excludes nz_specific). RT_391 stays open. Adds the processing-by-product leg to the corpus’s existing NZ hemp oil picture (LIT_089, OT_168, OT_112, OT_169); existing NZ hemp parameters unchanged. LIT_098
  • Oilseed rape is NOT “industrial only” — the processing and market legs (OT_183, Pure Oil NZ Grower Guide). Verbatim: “a large commercial size oilseed Crush Plant located 20Km South of Christchurch which produces high grade cold pressed oil and meal from New Zealand grown GE Free oilseed rape”, using “a cold pressing process with no chemicals (solvents) or added heat”; grown “through forward contracts with farmers throughout New Zealand, on both irrigated and non-irrigated properties” at 860–900/tonne**; establishment target **55–65 plants/m²**, nitrogen **150–200 units** [[ot_183_pureoil-osr-grower-guide|OT_183]]. 🎯 **This is what refutes dismissing oilseed rape as industrial:** the NZ commercial route is **cold press without solvents or heat** — the same broad method a community-scale press uses — and the crop is contracted **nationally on non-irrigated land**, partially offsetting [[lit_095_fasi-2012-osr-sowing-date|LIT_095]]'s irrigated trial (qualitatively; it supplies **no** unirrigated yield). ⚠ **No yield or oil-content figure appears anywhere in the guide** (whole-document search) — the two sources are complementary, neither is sufficient alone. ⚠ **860–900/t is content current DECEMBER 2020 — de-date before use; OT_128 is an input-cost index and is not a valid deflator for an oilseed price (no substitution made). 🔴 Regenerative tension, stated not smoothed: at 150–200 units of N, the only end-to-end NZ oil pathway is a higher-input, non-N-fixing crop — a live design conflict under D_001, comparable to hemp’s 100–130 kg N/ha (OT_112). OT_183

🔴 Oil-crop lane readiness — every NZ figure above is oil in the seed, not oil recovered

PathwayDefensible kg oil/ha?Why
Oilseed rape✅ the only one~1,008 kg oil/ha in the seed; both legs from one NZ trial (lit_095_fasi-2012-osr-sowing-date) + cold-press processing (ot_183_pureoil-osr-grower-guide)
Hazelnut⚠ blocked on a unitper-ha yields exist (ot_182_hazelnut-growers-bulletin1) and oil % exists (lit_096_savage-1998-nz-hazelnut), but the yields don’t declare in-shell vs kernel while the oil % is kernel-basis — multiplying could overstate ~2× (RT_393)
Olive❌ nokg/tree only; no NZ source publishes trees/ha (RT_392). Usable as kg oil per tree (~1.5 kg 2021, ~0.75 kg 2024), and only as a range
Walnut❌ nooil content only (lit_097_zwarts-1999-nz-walnut); no NZ yield at any scale
Hemp➖ unchangedexisting NZ params stand (lit_089_townshend-2010-fasamo-hemp-canterbury, ot_168_nzhia-hemp-production-2003)

🔴 Cross-cutting blocker: the recovery fraction is UNSOURCED (RT_391). The general primary (Singh & Bargale 2000, J. Food Engineering) is not held; the only recovery evidence held is hemp-specific and non-NZ (LIT_098). Therefore the upstream research report’s “806 kg recovered oil/ha” rests on an unverified 80% and must not be adopted — report oil-in-seed with an explicit flagged recovery band instead. The report’s “73%→80% recovery” phrasing also does not match snippets of the abstract it cited, suggesting a paraphrase error.

⚠ A pattern worth naming: NZ tree-crop industry publications routinely omit the denominator — the olive census never prints “hectare”, the hazelnut bulletin never declares in-shell vs kernel, and its Appleby 4 t/ha never states its spacing.

NZ community food self-sufficiency — case evidence

  • A CLAIM REGISTER, not an evidence base (CR_055, contested: true). An AI research synthesis on NZ small-community food self-sufficiency (dairy, edible fats, case evidence) for a ~30-household / ~80-person community CR_055. Every retrievable primary behind its headline numbers has been independently retrieved and verified — they, not it, are what the model cites. 🔴 Verification found a 3× error (98.26 reported as PEOPLE per day is in fact SERVINGS per day; the source’s own table reads “People per day … 33” — OT_193) plus eight further mis-attributions or uncited assumptions, each corrected on the named page: the uncited 90% milkfat recovery and the FAO cost-manual scope error (OT_173), the unverified 80% press recovery behind “806 kg recovered oil/ha” (LIT_095), hazelnut cold-hardiness figures from the wrong document (OT_182), a “4-day use-by”/“70°C” attributed to regulations that contain neither (REG_031), and a twice-a-day yield mislabelled as a house-cow figure (LIT_090). ✅ Its three declared NEGATIVES held: no NZ tallow/lard per carcass (its own instruction “Do not ingest as NZ coefficient yet” — RT_394), no verified NZ sunflower seed yield, and no NZ community source quantifying diet share / land per person / cooking-fat strategy (⚠ narrowed at ingest: URL_030 does run an olive pathway, it is just unquantified). Operating rule: where the synthesis and a retrieved primary disagree, the primary wins; where its primary was never retrieved, the figure does not exist. No figure on that page is model-eligible. CR_055
  • ⭐⭐ The ONLY NZ whole-diet land coefficient — and why 81% of it is meat and dairy (OT_190, Millar & Bould 2015, Blueskin & Karitane foodshed, contested: true). A foodshed of ~2,800 people; average NZ omnivore diet 826.1 kg/person/yr (veg 199.4 · dairy 117.2 · fruit 109.3 · beverages 92.6 · grains 90.9 · meat 66.0 · poultry 34.4 · fish 26.2 · eggs 9.3 · other 80.8); land for full food self-sufficiency 1,375.7 ha = “just 9%” of the foodshed’s 16,084 ha, or 16% of its pastoral land → ~0.49 ha/person, of which meat 0.287 + dairy 0.110 = 81% OT_190. 🎯 The structural finding survives all three defects: a community’s food land requirement is dominated by ANIMAL PRODUCTS, not the vegetable garden — the opposite of where the corpus’s evidence has concentrated. Also defect-free and model-relevant: land is not the binding constraint (9% of the foodshed suffices — the NZ empirical counterpart to LIT_012’s Auroville finding); the measured mass balance (four-legged meat 0.14, dairy 0.62, eggs 5.54, overall 1.26, with zero commercial production of fruit, vegetables, grains or beverages); measured foodshed dairy (5 farms, 532.5 ha, 4,295 L/cow/yr, 526.29 t MS; model farm 182 ha = 152 platform + 30 runoff; Dunedin 2.72 cows/ha); and the household survey — 65 m²/person veg space on residential sections vs 25 m²/person on large lifestyle blocks, “Not one household was completely self-sufficient in all of the food types”, only 8% self-sufficient in two or more. 🔴 THREE DEFECTS travel with the 0.49: (1) the yields are not measured — Table 12 is “modified ecological footprint data taken from… (Lawton, 2012)” and the report’s own measured meat yield is 165 kg/ha against Table 12’s 230 kg/ha; (2) the dairy row does not reconcile — every other row = (kg/person ÷ kg/ha) × 2,800 exactly, dairy computes to 332.1 ha but prints 308.00; (3) probable unit mismatch — dairy product weight against a milksolids yield, which would make the dairy footprint (22%) materially overstated and 0.49 ha/person ~25% too high. Note the errors run in opposite directions and do not cancel knowably. ➜ Use 0.49 ha/person as a CROSS-CHECK, not a model input, until Lawton (2012) is retrieved (RT_395). ⚠ The community-financed Jersey dairy hub (unpasteurised milk/cream/yoghurt, capitalised by customer advance purchase) is a 2015 proposal, not an operating business — and its distribution model runs straight into REG_031 reg 64. OT_190
  • The correction source — a 3× error caught by opening the primary (OT_193, UC GEOG402 2013, data_quality: low). The source’s own Table 4 reads “In 192 Beds … 98.26” (servings per day) and “People per day … 33”, with the text “a 0.4 acre parcel of land can therefore feed, on average, 33 people per day” OT_193. 🔴 CR_055 reports the 98.26 as PEOPLE per day. It is SERVINGS — a 3× overstatement of vegetable-bed carrying capacity which, had it been applied, would have understated the vegetable land requirement threefold. ⚠ low deliberately: student report, not peer-reviewed, and its yields are US data (Relf 2009; Garden of Eden Project 2013) with US/Canadian bed geometry (SPIN Farming) applied to a Christchurch site. Weak triangulation: its figures imply ~49 m²/person of parcel, bracketing OT_190’s derived 53.1 and its measured 25–65 m²/person survey range — but two of those four figures are verification-pass derivations, not printed values, so it is a sanity band, not a coefficient. 🔴 Vegetables only — never a whole-diet coefficient (that is OT_190’s ~0.49 ha/person, 81% animal products; confusing the two is a two-order-of-magnitude error). 🎯 A pattern in the corrections: four of the most consequential findings are UNIT errors, not data gaps — servings-vs-people here, in-shell-vs-kernel (OT_182), milkfat-vs-milksolids (OT_172), and product-weight-vs-milksolids (OT_190). OT_193
  • The one NZ community actually pressing its own oil — unquantified (URL_030, Awaawaroa Bay, Waiheke; data_quality: low). 🟢 Verbatim: “We have numerous olive trees that we harvest and press – producing some of the best olive oil around!” — an operating NZ community grow-harvest-press loop URL_030. This narrows CR_055’s third declared negative: it is not that no NZ community operates an oil pathway, but that no NZ community source quantifies one. ⚠ No tree count, no yield, no volume — and with no NZ planting density (RT_392) it cannot be converted against the census figures (~1.5 kg oil/tree 2021, ~0.75 kg 2024). Land structure: 169 ha total, ~160 ha collectively owned, 85 ha covenanted forest, 15 residential shares, 1 ha private each of which ~6,000 m² is growing allotment; off-grid Victron solar; 35,000 L of water storage; households keep chickens, cows and sheep. 🎯 Half the holding is covenanted forest and cannot be farmed — with OT_192’s Rainbow Valley (only 26 of 103 ha flat), the lesson is that a community’s total hectares systematically overstates its food land. 🔻 Tenure is the binding constraint: “as it is collective land, it is very difficult to raise a mortgage, so buyers will need to have at least $400,000 deposit” — the second independent NZ instance in the corpus of collective tenure converting a mortgage market into a cash market (the first: multiply-owned Māori land, “the land cannot secure any loans”, OT_191). Across the case evidence, land tenure and finance are the observed binding constraints, not agronomy. 🔴 PROVENANCE CORRECTION: this is NOT the ecovillage’s official site — it is one household’s marketing page for the sale of their residential share, which CR_055 cited as the community’s site. ⚠ A for-sale page will disappear; the corpus holds a local HTML capture. URL_030

Nutrient loop & crop-N calibration (LIT_106–110)

  • Excreta-N as a fertiliser-nitrogen substitute closing the food nutrient loop (global proxy, Zheng et al. 2024). Globally, human excreta carried 31.8 Mt N/yr in 2020 of which 23.3 Mt/yr is recoverable after sanitary treatment — the source behind the engine’s per-person recoverable-N constant that offsets the food system’s crop fertiliser-N demand when a composting toilet is used, framing on-site excreta recovery as displacing imported synthetic N (the food-side half of the D03 wastewater→food loop). ⚠ The EXCRETA_N_RECOVERABLE_KG_PERSON_YR = 3.0 constant is DERIVED (23.3 Mt ÷ ~7.8 bn ≈ 2.99), not a per-capita value in the paper; corroborated by independent eco-san estimates (~4 kg N excreted/person/yr). Abstract-only (paywalled); global aggregate, no NZ figure. LIT_106
  • Potato nitrogen demand — the primary behind CROP_N_DEMAND_KG_HA["potato"] = 110 (LIT_107, INTERNATIONAL PROXY). Varga et al. (2025) peer-reviewed review (MDPI Nitrogen 6(4):117): potato average requirement 80–120 kg N/ha, recommended optimal 100–120 kg N/ha for high yield with preserved dry matter/starch, >150 kg N/ha counterproductive (delayed maturation, lower dry matter/starch, wasted fertiliser, greater N loss). The engine’s potato N constant of 110 is the midpoint of the recommended band. Strongly cultivar-dependent (early cultivars saturate ~100; late e.g. Daytona/Ninfa/Spunta respond to ~200) and region-dependent (high-yield systems optimize 150–250) — so 110 is the balanced-quality planning value, not a universal maximum. ⚠ Croatia/EU study, transferable proxy, no NZ value (same convention as LIT_060). LIT_107
  • Vegetable N management — the general reference, and a correction on the veg N-demand cell (Tei et al. 2020, Agric. Water Manag. review). Vegetable crops have low N-use efficiency and a high but strongly species-specific N demand; this qualitative review gives no single mixed-vegetable N-rate — it defers per-crop rates to look-up tables (Feller & Fink 2002 Nmin target values, Acta Hortic. 571; AHDB RB209 §6). Its only quantitative veg-N anchor is crop-residue N: 25–30 kg N/ha (spinach, lettuce) to 250–300 kg N/ha (cabbages), 60–80% mineralised within 3 weeks. 🔴 Correction: does NOT support CROP_N_DEMAND_KG_HA["veg"]=150, and its “151 kg N/ha” is soil N-mineralization potential, not crop demand — re-source that cell from RB209 / Feller & Fink or carry a range. International (EU) review, transferable proxy, no NZ value. LIT_108
  • Excreta-P as a food-self-sufficiency lever (France, transferable proxy — no NZ value). Recycling all human-excreta P could cover 7-34% of the P in French food supply without changing the food system (7% current diet; 34% ceiling when excreta-P is directed to plant-based human-food crops), because plant products have a far higher Phosphorus Use Efficiency than animal products. A recognised, quantified precedent for treating recovered excreta nutrients as a measurable share of a community’s food-P need. ⚠ P only; a food-supply-scope metric, not a fertiliser-demand offset (that is ~15% directly). LIT_109
  • Recovered urine offsets bought synthetic fertiliser (transferable proxy, no NZ value). The same review finds urine-derived fertilisers “can match or exceed mineral fertilizers, with yield improvements of 10-70% compared to unfertilized controls and yields comparable to urea when applied at equivalent N rates” — the plant-level evidence that recovered urine-N can substitute for purchased synthetic N in a community food system (the D02↔D03 nutrient loop). ⚠ International review; transferable proxy only. LIT_110

Community cool-store sizing

  • Community cool-store sizing — crop shelf-life, design ratio and refrigeration load. Post-harvest storage life by crop sets how much of the year on-site produce can cover: broccoli 2–3 weeks, cabbage 12–16 weeks, apple 6–12 months, potato ~12 months (USDA Agriculture Handbook 66 OT_209; Cornell cold-storage chart OT_210), with carrots measured to ~6 months in long-term storage OT_208. Size the room from a gross ≈ 1.5× net design ratio (usable produce volume × ~1.5 for air-gaps, stacking and circulation — FAO’s factor, drawn from a meat cold-store manual) OT_207, and budget refrigeration at ~10–14 kW per 1,000 m³ of room (USDA HB66) OT_209. ⚠ Open gap: converting stored tonnage → room volume needs a produce bulk-density (~200–250 kg/m³ assumed) for which no authoritative source is held, so the cool-store sizing synthesis is incomplete.

Producer cost + oil/fat recovery fractions

  • Screw-press oil recovery fraction — the cross-cutting oil-lane multiplier, now sourced (LIT_122; resolves RT_391). A small cold screw press recovers ~50–85% of the oil in the seed (hemp 53–81%), so usable oil = seed yield × oil-in-seed × recovery — the term the corpus’s oil-in-seed figures (LIT_095 rape, LIT_096 hazelnut) were missing; the OSR ~1,008 kg/ha must be de-rated by this band, not counted whole. ⚠ Transferable proxy (press physics), conservative small-press floor ~50%. LIT_122
  • Hazelnut crack-out fraction — resolves the in-shell-vs-kernel unit block (OT_211; resolves RT_393). Crack-out is ~46% (kernel ÷ in-shell) and NZ per-ha hazelnut yields are stated in-shell, so apply the ~46% crack-out before the kernel-basis oil content (LIT_096 54.6–63.2%), removing the ~2× overstatement risk against OT_182’s t/ha yields. OT_211
  • Grazing establishment CAPEX — the fencing + water legs of the livestock cost cell (OT_212 / OT_213; with the OT_091 OPEX leg, resolves RT_280). Stock-water reticulation ~200/ha (2026)** ([[ot_212_blnz-stock-water-reticulation|OT_212]]) and non-electric 8-wire fencing **~13–17/m (2015) (OT_213) — the fence rate validates the held **33.35/m** ([[cr_044_nz-livestock-capex-opex-costs|CR_044]]) as a defensible conservative 2026 steep/contractor rate. ⚠ Stockyards + stock-purchase /head legs remain unsourced. OT_212 OT_213
  • Small-scale butter fat-recovery fraction — the empty term in the milkfat→butter chain (OT_214; resolves RT_388). Direct sour-whole-milk churning (no separator) recovers 44–76% of milkfat (76% ceiling at 18 °C with an agitator churn); a separator + churn ~88% (to ~95%); the 0.82 butter fat content divisor is confirmed (OT_173) — correcting the uncited 90%, so butter_kg = milkfat_kg × recovery ÷ 0.82 is reported as a band across the sourced recovery range. OT_214
  • Lamb edible-fat coefficient — the animal-fat leg, NZ lamb tier (LIT_123; resolves RT_394 at the lamb tier). NZ pasture-finished lamb dissectible (separable) fat 13.7% (Romney; ¾ Wiltshire leaner at 11.2%); edible rendered tallow ≈ HCW × separable-fat% × ~0.78 (rendering recovery). ⚠ Leg-basis carcass proxy; the ~0.78 render fraction, national mean HCW, and beef ~20% / pig ~22% remain flagged proxies (no NZ carcass-fat figure held). LIT_123

Connections

Links to

Sources (109): CR_002 · CR_004 · CR_005 · CR_018 · CR_021 · CR_033 · CR_034 · CR_039 · CR_041 · CR_043 · CR_044 · CR_051 · +97 more

Technologies (3): Agro-PV (Dual-Use Solar + Agriculture) · Biogas (community-scale,… · Biodynamic Farming

Decisions (1): D_001: Neobiome food production — regenerative…

Concepts (1): EDT Framework — Emerging & Disruptive Technolo…

Cases (1): Electric Cherries

Referenced by

Sources (36): CR_051 · CR_052 · CR_054 · CR_058 · LIT_076 · LIT_088 · LIT_109 · LIT_115 · LIT_122 · LIT_130 · OT_091 · OT_096 · +24 more

Concepts (1): Low-Carbon & Bio-Based Construction

EDT domains (1): D01: Renewable Energy & Storage Systems

Graph View

Backlinks