Research Brief — R07: Soil vs. Coco vs. Hydroponic Systems in Cannabis Cultivation

Report ID: R07 (Wave 2, systems comparison; executed eighth — closes Wave 2) · Brief version: v1.0 — FROZEN September 9, 2026 on Danny’s confirmation (ChatGPT final look on the full v0.96 text: APPROVED FOR FREEZE, no further changes). Both research runs execute this version and must cite it; the critique stage records it. (v0.9 drafted 2026-09-09. v0.95 same day after ChatGPT’s preliminary review — seven calibrations accepted: substrate and delivery architecture declared orthogonal axes; “hydroponic” pre-registered as a taxonomy before search; genotype × system and environment × system interaction warnings; cycle-time reporting chain; eleventh field expanded with drainage, geometry, fertigation frequency and supply mode; economics normalized to CAPEX / OPEX / labor / output-per-time; water and nutrient metrics enumerated; system-property matrix adopted as the primary synthesis target — and Gemini’s review of the Drive doc, four technical items taken up: coir cation exchange, dissolved oxygen vs. temperature, coir particle grades, named research groups. v0.96 same day after ChatGPT’s disposition on Gemini’s items: the three technical prescriptions reframed as research targets rather than stipulated answers — coir CEC / washing-vs-buffering / exchangeable K, Na, Ca, Mg to be characterized, not asserted; the universal “20–22 °C trigger” and “primary cause” language removed in favor of temperature × dissolved oxygen × biological oxygen demand × pathogen pressure as an interaction to investigate; the hard-coded air-filled-porosity range removed in favor of a measurement requirement; literature leads kept as search targets with citation-network expansion and competing literature explicitly allowed; root-zone temperature distinguished from air temperature added to the methodology and the eleventh field. Gemini’s numbering complaint remains rejected on the text: questions are numbered continuously 1–28 across A–H, the R06 convention, so B3 and H24–H27 are correct as written — ChatGPT’s conditional endorsement of that fix rested on Gemini’s description, not the document.) Date drafted: September 9, 2026 · Drafted by: Claude (editor) · Protocol: v1.4 (with the §3 standing quantitative-claim fields adopted 2026-09-07 from R04)

Methodological spine (one sentence): name the root-zone system by its measured properties and architecture — medium composition and physics (air-filled porosity, water-holding capacity, cation-exchange capacity, pH buffering), container or bed volume, delivery architecture (hand-water / drip drain-to-waste / recirculating / flood-and-drain / deep-water / NFT / aeroponic), irrigation frequency and leach fraction, and root-zone temperature — never by the labels “soil,” “coco,” or “hydro” → hold the nutrient program, environment and genotype fixed or name the mismatch → keep the three outcomes distinct (cannabinoid concentration, flower yield, total cannabinoid yield) and add the time axis (yield per m² per cycle and per year) → compare only on a stated functional unit and boundary → attribute effects to root-zone oxygen, water availability, buffering, nutrient dynamics and pathogen ecology, never to the category.

Assigned to: ChatGPT deep research + Gemini deep research (independently, same frozen brief; both runs cite the frozen version)

Assignment

Produce an evidence review of where cannabis roots live — mineral and organic soils and soil-based potting mixes (in ground, beds, containers); soilless organic substrates (peat-based mixes, coconut coir and coir blends, bark and wood-fibre composites); inert substrates (rockwool/stone wool, perlite, clay pebbles/LECA, foam); and water-culture systems (deep-water culture, nutrient-film technique, aeroponics, flood-and-drain over inert media) — and what the record establishes about their effects on growth rate and cycle time, yield, cannabinoid and terpene chemistry, plant health and pathogen risk, water and nutrient use, waste and end-of-life, labor and skill demand, and environmental footprint.

This is not a “which is best” report (PROTOCOL research voice). It is a map of what has been measured in cannabis, what is borrowed from controlled-environment horticulture, and what is asserted — with the mechanisms that explain differences where differences exist.

R07 closes Wave 2 and must inherit, not repeat:

• R06 (fertility): the license — cultivation outcomes are properties of nutrient systems, of which substrate, root-zone chemistry, irrigation water and delivery architecture are named components (R06-C001); fertigation architecture is an independent variable — recirculating vs. drain-to-waste did not conclusively change inflorescence mass but significantly raised total THC yield in one trial (R06-C015, S-205); solution EC is an operational control, not an equivalence measure (R06-C021); organic-fertilizer optima were substrate-dependent in coir blends (R06-C008, S-100/S-101); flushing/leaching regime effects differ by duration (R06-C024/C025) and a drain-to-waste flush is a discharge event (R06-C027); microbial compliance is assay-dependent (R06-C032). R07 takes over root-zone architecture, medium physics and water/waste flows as its own; nutrient chemistry stays R06’s.
• R08 (living soil): the keystone absence — no controlled, replicated, multi-cycle study compares persistent indoor beds against conventional substrate systems on yield/area/time, chemistry, sanitation, labor or economics (R08-C030); the one controlled indoor organic-vs-mineral comparison (R08-C006, extended as R06-C006); media reuse as an engineered practice with adjacent-CEA evidence for peat and coir (R08-C013/C014/C015, S-110–S-113); root pathogens recovered from coir and rockwool in commercial production (R08-C017, S-116); long-term container pH driven by irrigation-water alkalinity more than by medium (R08-C024); media as a minor indoor carbon term (R08-C027). R07 extends the media-reuse and pathogen-ecology threads to soil and inert media and to water culture.
• R04 (environmental footprint): the indoor water balance — ~70% of applied water recovered as HVAC condensate in the one metered study, no representative commercial water balance located (R04-C018, S-007); media as the dominant outdoor-container carbon term and a substantial modeled reuse lever (R04-C030, S-003); no mass-flow inventory of growing media, rockwool or plastics located — “hydroponics leads to a lot of waste” is an unquantified assertion in both the archive and the literature (R04-C033); the functional-unit reversal — fewer plants, less media, lower per-kg footprint (R04-C029); “organic/living-soil has a lower footprint” left open as R04 M10 and R06 M10.
• R05 (environments): functional-unit sensitivity (R05-C032) and non-averageable lifecycle boundaries (R05-C033); heavy-metal risk tracks growing media, soil history, fertilizers and water rather than environment (R05-C024); supplemental light raises evapotranspiration but lowers ET per gram (R05-C015).
• R01 (organic agriculture): the functional-unit rule (R01-C018) and the practices-not-labels license (R01-C033) — a root-zone system is a bundle of practices, and the label is not the mechanism.

Tag every load-bearing claim by evidence type (PROTOCOL §3 standing fields): measured / hybrid / modeled / observed-administrative / regulatory / absence / methodological. For R07 the dominant types will be measured trials (mostly adjacent CEA), a small number of cannabis trials, and absences; label each.

Methodology requirements

Systems are named by their measured properties and architecture, never by label (the spine). “Soil,” “coco,” and “hydro” are trade categories that each cover incompatible systems: “soil” spans native field soil, amended beds, and bagged peat-based “soil” mixes that are soilless by any physical definition; “coco” spans buffered and unbuffered coir, pith/chip/fibre ratios, and coir–perlite blends with different air-filled porosity; “hydro” spans inert-media drip systems whose root-zone physics resemble coir more than they resemble deep-water culture. For every system discussed, record: medium composition and source; physical properties (air-filled porosity, container capacity / water-holding, bulk density, particle-size distribution where reported); chemical properties (cation-exchange capacity, pH buffering, initial EC, sodium/chloride load in coir); container or bed volume and root-zone depth; delivery architecture (hand-water, drip drain-to-waste, recirculating, flood-and-drain, DWC, NFT, aeroponic); irrigation frequency, leach fraction and dry-back regime; root-zone temperature and dissolved oxygen where reported; and whether the nutrient program was held constant across systems. Where a source does not disclose these, say so and grade the claim accordingly. A yield or quality figure from an undescribed “soil” or “hydro” system is estimate-lore.

Substrate and delivery architecture are orthogonal axes — never infer one from the other. Coco is not drain-to-waste; rockwool is not “hydroponics generally”; a soil-based mix can be drip-fertigated almost like an inert substrate; DWC, NFT and aeroponics are radically different root environments from containerized coir even though all four are “hydro.” Every system in the inventory is placed on both axes independently — (substrate class) × (delivery architecture) — and a result observed on one axis is never attributed to the other without a design that separated them.

“Hydroponic” is pre-registered before the search. The literature uses the word for at least four different things: (a) soilless cultivation broadly, including peat and coir; (b) any nutrient-solution feeding without soil; (c) recirculating water culture specifically; (d) inert-substrate drip fertigation. For R07: “water culture” = roots in solution or mist with no solid substrate (DWC/RDWC, NFT, aeroponics); “inert-substrate fertigation” = stone wool, perlite, LECA, foam under drip, flood-and-drain or hand-water; “organic soilless substrate” = peat, coir, bark and wood-fibre mixes; “soil” = native soil, amended beds, or mixes containing mineral soil. When a source says “hydroponic,” the review records which of (a)–(d) it means and re-files the system under the R07 class; a source that cannot be re-filed is taxonomy-undisclosed. Both researchers use this taxonomy so that nominally similar evidence is not retrieved for different systems.

Genotype × system and environment × system interaction. A cultivar that performs well under high-frequency inert-substrate fertigation may not rank the same in a buffered organic substrate or a bed, and indoor climate management (VPD, light intensity, root-zone temperature) can interact strongly with root-zone water and oxygen behavior. Single-genotype trials are labeled as such wherever they are cited and synthesis conclusions drawn from them carry the label; multi-genotype or multi-environment trials report whether an interaction was tested. No system ranking is generalized across genotypes or climates without that test.

System-comparison equivalence is the comparison rule. No two root-zone systems may be compared on any outcome unless the review states whether they were matched on (i) nutrient program — chemistry, rate and availability schedule (R06 discipline; equal solution EC is not equal nutrition, R06-C021), (ii) environment (light, CO₂, VPD, root-zone temperature), (iii) genotype, (iv) container or root volume and plant density, (v) duration and harvest criterion, and (vi) irrigation regime (frequency, leach fraction). A trial that changed medium and fertilizer and irrigation frequency together — as most practical “soil vs. hydro” comparisons do — is a system-bundle comparison, not a medium comparison; say so, and never attribute the result to the medium alone. Where treatments were not matched, report the result with the mismatch named.

Three outcomes plus the time axis. Cannabinoid concentration, flower yield, and total cannabinoid yield stay distinct (the R08-C006 / R06-C006 lesson). R07 adds the dimension that system comparisons most often invoke: cycle time. “Hydro grows faster” is a claim about days to harvest, not grams; yield per m² per cycle and per m² per year are different functional units and can rank systems differently. Every growth-rate claim names the metric (days in veg to a defined size, days to harvest, biomass accumulation rate) and the harvest criterion; every yield claim names per-plant, per-m², per-cycle, per-year, or per-kWh, and states the plant density. Where a source allows it, the review reports the full cycle-time chain: vegetative duration (and the size criterion that ended it) → flowering duration (and the harvest criterion) → harvest dry mass → area-and-time basis. A shorter cycle is economically meaningful even when yield per crop is unchanged; the two are never collapsed into one “faster/better” verdict.

Functional-unit and boundary discipline (R01-C018, R05-C032, R05-C033, R04 spine). Per-plant, per-m², per-cycle, per-year, per-kg-flower, per-g-cannabinoid, per-litre-water and per-kWh are not interchangeable. Environmental comparisons state both per-area and per-product framings where both exist; R04-C029 remains the standing example of a per-kg ordering that reverses the per-plant intuition, and R04-C030 the standing example of media as the dominant term in one boundary and a minor term in another.

Water accounting names the flow. “Uses X% less water” is uninterpretable without the boundary: applied irrigation, transpired, drained to waste, recirculated, evaporated, recovered as condensate, RO reject, and discharged. Every water claim names which flows are counted and whether the figure is measured, metered at facility level, or modeled; recirculating-system water savings are reported against the drain-to-waste baseline actually used, not against a field-irrigation figure (R04-C009’s descent from an ordinance assumption is the cautionary case).

Root-zone temperature is not air temperature. Two rooms at the same air temperature can hold materially different root-zone thermal regimes — a chilled or unchilled RDWC reservoir, a black pot of coir under HPS, a stone-wool slab on a heated bench, a large soil bed with thermal mass — and root-zone temperature influences oxygen availability, root metabolism, microbial activity, irrigation demand and disease pressure. Every trial in the inventory records root-zone temperature where measured and is flagged air-temp-only where only ambient is reported; system comparisons that differ in root-zone temperature name the difference.

Irrigation-water chemistry is a comparison field (inherited from R06). Source water, alkalinity, starting EC, Ca/Mg and Na/Cl alter pH behavior and buffering enough that a reported medium effect can be a water-medium interaction (R08-C024). Every trial in the inventory records irrigation-water chemistry where disclosed and is flagged water-undisclosed where not.

Medium physics before medium label. Where systems differ in outcome, the review attributes the difference to a named mechanism with its own evidence — root-zone oxygen and air-filled porosity; water availability and matric potential vs. dry-back; nutrient buffering and CEC; pH drift; root-zone temperature; root morphology and root:shoot allocation; salt accumulation; pathogen inoculum and suppressiveness — and grades the mechanism separately from the outcome. Mechanisms borrowed from tomato, cucumber, leafy greens, ornamentals or strawberry are adjacent; from general plant or soil science general; never direct.

Pathogen ecology is a two-sided question. “Sterile” inert media and water culture remove soil-borne inoculum but also remove suppressive microbiology; root-zone pathogens have been recovered from coir and rockwool in commercial cannabis (R08-C017). Every disease claim names the pathogen, the medium, the system architecture (recirculation spreads waterborne oomycetes), the detection method, and whether the comparison was controlled.

Practice–effect separation. Split each practice (dry-back cycling, high-frequency fertigation, leach-fraction targets, “crop steering,” media buffering, media reuse and sterilization, reservoir management) from its claimed effect and grade each separately. Adoption is evidence about the practice, not the effect. Unmeasured effects stay unverified — a status, not a grade of weakness.

Safety claims name the compound, the medium, and the assay. Heavy metals in coir (source geography, processing), peat, biosolids-derived mixes, and field soils with site history (R05-C024); sodium/chloride in unbuffered coir; microbial loads from soil vs. sterile media (R06-C032 assay dependence); mineral fibre and dust from stone wool as a worker-safety question. Every safety claim names the compound or organism, its source, the matrix tested, the jurisdictional limit, and the assay. Regulatory claims carry jurisdiction and as-of date.

Evidence type and evidence quality, graded separately (R04 standard). Type tags per §3. Quality on four axes: relevance (does it answer the question on the system, outcome and functional unit asked?), representativeness (genotype, scale, cycle count, single site?), transparency (can the system be reconstructed from the paper?), and reported uncertainty (replication, variance, statistics). Opacity disqualifies.

Provenance for quantitative estimates — 11 fields for R07. Every major quantitative result carries the standard 10-field record (geography, period, population/genotype, metric, functional unit, system/medium, boundary, method, directional bias, confidence) plus an eleventh field for this report: “root-zone system descriptor” — (1) substrate composition and source; (2) container or bed geometry (volume, depth, footprint, plant density); (3) drainage characteristics (free-draining container, lined bed, in-ground; leach fraction); (4) delivery architecture (hand-water / drip / flood-and-drain / DWC / NFT / aeroponic) and recirculating vs. drain-to-waste status; (5) fertigation or irrigation frequency and dry-back regime; (6) measured or reported water-holding and air-filled porosity where available; (7) nutrient-supply mode — solution-driven (soluble feed at each irrigation) vs. substantially mineralization-dependent (pre-amended or top-dressed organic sources) vs. mixed; (8) root-zone temperature (solution, substrate or bed — measured, not inferred from air temperature; flag air-temp-only where the source reports only ambient) and dissolved oxygen where reported; (9) irrigation-water chemistry; (10) whether the nutrient program was held constant across systems (with the R06 equivalence basis where it was not). The descriptor is what makes a claim retrievable by property rather than by the words “coco” or “soil.” The field is conditional: for a single-system study it reads “not applicable — single system, descriptor: […]”. A blank eleventh field on any cross-system comparison means the comparison is uninterpretable and is labeled system-orphaned.

Jurisdiction- and date-stamping on every regulatory and compliance claim. Prefer primary sources: peer-reviewed controlled trials in cannabis and hemp; then controlled-environment horticulture (tomato, cucumber, pepper, leafy greens, strawberry, ornamentals) as adjacent; substrate-science literature (peat, coir, stone wool physics) as adjacent or general; state and federal agency documents; utility and extension trials; then industry white papers and journalism, labeled as such.

Key questions

A. Definitions, taxonomy, and physical properties (required output category)

1. Build a root-zone system taxonomy table: for each system class used in cannabis (native soil in ground; amended raised or indoor beds; soil-based container mixes; peat-based soilless mixes; coir and coir blends; bark/wood-fibre composites; stone wool; perlite, LECA and other inert aggregates; flood-and-drain; DWC/RDWC; NFT; aeroponics) — typical composition, air-filled porosity and water-holding ranges, CEC and buffering, initial pH/EC and known contaminants (coir Na/Cl, peat acidity), typical container volume and delivery architecture, and what is unknown or highly variable about each. Coir is not one substrate: pith, chips and fibre, their blends, compression history and container geometry all change water retention, drainage and air-filled pore space. Every coir entry names the grade or blend (or is flagged grade-undisclosed) and reports measured air-filled porosity and water-holding where the source gives them; the review does not assume a porosity range for a grade it has not seen measured. Coir chemistry is a research target, not a stipulated answer: characterize coir cation-exchange capacity, the difference between washing (removing soluble salts) and buffering (treating exchange sites with Ca/Mg), and the roles of exchangeable K, Na, Ca and Mg in what growers report as “coco problems” — and distinguish those exchange-site mechanisms from bulk soluble-salt / EC effects. State what has been measured in cannabis, what is adjacent substrate science, and what is vendor assertion.
2. Which physical and chemical substrate properties have been measured in cannabis-specific studies versus borrowed from horticultural substrate science? Inventory the cannabis substrate-characterization literature (e.g., the Nemati review, S-134; Ahmed 2023, S-096; Caplan substrate work, S-100/S-101) and state what it covers and what it does not. Named search targets (groups, not pre-verified citations — every paper located still passes identity verification): the University of Guelph controlled-environment group (Zheng and co-authors: substrate, fertigation and irrigation work in cannabis, including the Caplan and Stemeroff theses already registered); the Volcani Institute group (Bernstein and co-authors, including Shiponi and Danziger: root-zone nutrition and any root-restriction or aeration work); and the Guelph aquaponic/water-culture cannabis papers (Yep and co-authors). These are leads, not privileged authorities: search them explicitly, then expand through their citation networks and into competing or contradicting literature, and report which of these groups’ cannabis papers bear on root-zone systems as opposed to nutrition alone.

A practitioner-record passage is withheld from the public copy.

A practitioner-record passage is withheld from the public copy.

D. Quality, chemistry, and sensory

Sensory evidence comes in four classes that never collapse into “tastes better” (inherited from R06): (a) objective chemistry, (b) trained-panel sensory testing, (c) consumer-preference testing (blinded or not — say which), (d) grower or consumer anecdote. Every quality claim names its class. 10. Cannabinoid and terpene responses to root-zone system at fixed nutrition and environment: what is measured (which compounds, concentration vs. total), and does any study link medium or architecture to a sensory outcome — and in which class? 11. “Soil-grown has more terpenes / tastes better,” “hydro is bland,” and “coco gives cleaner terps” are three separate hypotheses, each with four possible evidence classes — not one generic “quality” outcome. For each: is there any class-(a) chemistry (which compounds, concentration vs. total), class-(b) trained-panel, or class-(c) consumer-preference evidence, or is it entirely class (d)? R07 is expected to be thinner here than R06; say so rather than borrow.

E. Plant health, pathogens, and safety

12. Root-zone disease by system: extend R08-C017 (S-116) with any cannabis or hemp evidence on Pythium/Globisporangium, Fusarium, and root-rot incidence by medium and architecture; the recirculation-spread question; the temperature × dissolved-oxygen × oxygen-demand interaction in water culture (C8), multicausal and not pre-registered as a threshold; suppressive vs. conducive media; and what detection method each study used.
13. Heavy metals and contaminants by medium: coir (source and processing), peat, biosolids-containing mixes, field soils with site history (R05-C024) — any cannabis measurement tracing medium contaminant to flower residue or compliance failure? Contrast with R06’s input-metal findings so the two reports do not double-count.
14. Microbial compliance by system (R06-C032 assay dependence): does any evidence show TYMC or pathogen-qPCR outcomes differ between soil, soilless and water-culture cannabis?
15. Worker safety: stone-wool fibre and dust, peat dust, and recirculating-reservoir biofilms — what is documented in cannabis or adjacent horticulture?

F. Water, nutrient, waste, and footprint flows

16. Water balance by system. Extend R04-C018 with any measured or metered water use by root-zone system in cannabis. These are different metrics and are never pooled: water applied per irrigation event; water applied per plant per day; water applied per m² per cycle; water withdrawn per kg dry flower (the R04/R05 functional-unit framing); fraction transpired vs. drained vs. recirculated vs. condensate-recovered vs. discharged. Each claim names which one it reports and whether it is measured, metered at facility level, or modeled; where only CEA-crop data exist, say so; evaluate “hydroponics uses 90% less water” against its actual baseline and metric.
17. Nutrient discharge by architecture. Extend R06-C027 and R04-C028. Again distinct metrics: discharge concentration (mg/L N, P, K in leachate or dumped reservoir); nutrient mass discharged per plant, per m² per cycle, and per kg flower; fraction of applied N, P and K leaving the root zone. Report each under drain-to-waste vs. recirculating vs. beds — measured or modeled — and what is known about cannabis nutrient-discharge regulation (jurisdiction- and date-stamped).
18. Media end-of-life and mass flow. Extend R04-C033: any inventory of spent media (rockwool, coir, peat, soil) generated per unit output; landfill vs. compost vs. reuse fates; stone-wool recycling programs and their documented uptake; the archive’s “hydro leads to a lot of waste” claim evaluated on magnitude, not direction.
19. Media reuse across systems. Extend R08-C013/C014/C015 (peat, coir) to soil beds and inert media: what evidence exists for multi-cycle reuse outcomes (yield, pathogen carryover, salt accumulation, physical breakdown), and what does reuse require (steaming, washing, re-amending, dry-back)?
20. Embodied footprint of media and systems on compatible boundaries (R04 discipline): peat extraction, coir processing and transport, stone-wool manufacture, perlite expansion, plastics and pumps for water culture; media as dominant outdoor-container term vs. minor indoor term (R04-C030, R08-C027); any LCA that varied medium at fixed environment.

A practitioner-record passage is withheld from the public copy.

H. Synthesis targets

The report does not end with a winner. It ends with a system-property matrix. The synthesis maps root-zone systems along measured or documented property axes — buffering capacity (pH, EC, water); root-zone oxygen control; nutrient-supply precision and speed of adjustment; water-use control; recirculation capability; biological dependence (mineralization, suppressive microbiology); media longevity and reuse; monitoring burden; failure velocity (time-to-symptom, time-to-crop-loss); automation compatibility; and capital and labor intensity. “Coco drip drain-to-waste,” “living-soil bed,” “rockwool drip,” “DWC,” “peat-mix hand-water” and the rest are then examples occupying different regions of that space, not contestants. The question R07 asks of the folklore — “soil is forgiving,” “hydro is fast,” “coco is the best of both worlds” — is which physical and operational properties produce the trait the slogan is describing, and whether the literature has been attributing to the medium’s name what was actually irrigation frequency, oxygenation, EC dynamics, buffering, nutrient availability schedule, or crop-cycle management (the R06-C001 and R08-C030 lesson joined). 24. The root-zone system map: every load-bearing result placed on its eleventh-field descriptor, nutrient-equivalence status, functional unit and time axis — the artifact that makes “soil vs. hydro” results non-poolable by construction — and the system-property matrix built from it, with each cell graded by evidence type and marked direct / adjacent / general / unmeasured. 25. The three-outcome-plus-time map (concentration / yield / total cannabinoid yield / cycle time) by system class, with named gaps. Absence of measurement is a finding: do not build a verdict on “soil vs. hydro quality” because readers expect one. 26. The flow map (water, nutrient discharge, media mass) by architecture, with boundaries named. 27. The mechanism table: each documented difference → named mechanism → evidence grade in cannabis vs. adjacent crops. 28. The 8–10 most repeated public claims and what the evidence says (M1–M10 below).

Pre-registered public claims to test

Word all verdicts so absence of evidence remains absence of evidence (“unsupported” ≠ “the opposite is true”; unknown is acceptable):

• M1. “Hydro grows faster and yields more than soil” — which metric (days, g/plant, g/m²/cycle, g/m²/year), at what density, and was nutrition matched? Distinguish cycle-time from yield-per-cycle from yield-per-year.
• M2. “Soil-grown cannabis tastes better / has more terpenes” — which of the four sensory classes has evidence, and which compounds would carry it?
• M3. “Coco is the best of both worlds” — what measured property of coir (AFP, water-holding, CEC, buffering) supports it, and against which soil and which hydro system?
• M5. “Hydroponics uses 90% less water” — against which baseline and boundary; is the figure cannabis-measured, CEA-borrowed, or marketing?
• M6. “Hydroponics leads to a lot of waste — rockwool goes to landfill” (archive, 2024 deck) — direction plausible, magnitude unmeasured (R04-C033); any mass-flow or fate data?
• M7. “Soil buffers mistakes; hydro shows problems in hours” — buffering and drift evidence by medium (C7), and whether faster detection is a cost or a benefit.
• M8. “You can’t grow organic in hydro / coco” — bioponic and organic-in-coir evidence (R06-C008 shows organic-liquid dose-response in coir); name the mechanism (mineralization requires biology and residence time).
• M9. “Soilless media are sterile, so fewer root diseases” — R08-C017 counters the premise; what is the controlled evidence on disease incidence by system and architecture?
• M10. “Living soil / soil has a lower environmental footprint than coco or hydro” — inherited from R04 M10 and R06 M10: per-area vs. per-product, media as boundary-dependent term (R04-C030 vs. R08-C027), water and discharge flows.

A practitioner-record passage is withheld from the public copy.

A practitioner-record passage is withheld from the public copy.

Standing source & citation requirements (PROTOCOL v1.4 — verbatim)

• Every citation supporting a substantive claim must carry a resolvable DOI, stable URL, or independently retrievable identifier — an unresolvable citation makes the claim unsourced.
• Citation-identity integrity: for every substantive source, the authors, title, publication year, venue, volume/pages where applicable, and DOI or stable identifier must resolve to the same bibliographic object — a DOI that resolves to a different real paper fails identity verification. If you cannot fully resolve the tuple, flag it explicitly as a partial citation rather than completing it from memory.
• Exact claim support: a source supporting the general direction of a finding does not automatically support its stated magnitude, population, geography, time period, boundary, or outcome measure — preserve these distinctions in candidate claims.
• Corrections inherit the complete verification burden: a replacement citation or corrected statistic remains unverified until its existence, identity, and claim support are checked.
• Source-identity deduplication and corroboration independence (rules 11–12): one paper is one source however many times it is cited; “multiple studies agree” requires demonstrably distinct studies or datasets.
• When testing public claims, “the claim is unsupported” and “the opposite is true” are different findings — do not overcorrect; unknown is an acceptable verdict.
• Standing quantitative-claim fields (§3, adopted 2026-09-07): every load-bearing claim carries an evidence-type tag (measured / hybrid / modeled / observed-administrative / regulatory / absence / methodological), and every quantitative claim carries an inline boundary line — system/medium · functional unit · outcome measure and method · population/genotype — with “N/A — reason” where a field does not apply. For R07, the boundary line also names the root-zone system descriptor and the time axis (per cycle / per year).
• Report ranges and disagreements between sources, not averaged-away conclusions; list data gaps explicitly.

Candidate claims for editorial review

Conclude with 15–30 concise, testable candidate claims with supporting citation(s), each carrying its system/medium/genotype/jurisdiction/period tags and, for magnitudes, the outcome measure, functional unit and time axis, root-zone system descriptor, nutrient-equivalence status, and evidence-type tag. Do not assign final evidence status or confidence; the editor will.

Search-stopping rationale

End with an explicit, bounded statement of why searching concluded: the search cutoff date, the databases and query strings covered, the inclusion criteria applied (language, publication types, date range, crop and system scope), saturation observed, and named question areas where no eligible studies were located within the documented search — distinguishing “not searched” from “searched and none located.” Do not write that the literature is “genuinely absent”; absence is a property of the search, not of the world.

What “done” looks like

A structured report answering A–H with cited, system- and outcome-tagged evidence; the root-zone system taxonomy (A1) and the controlled-trial inventory (B3) as distinct outputs; the root-zone system map and system-property matrix (H24), three-outcome-plus-time map (H25), flow map (H26) and mechanism table (H27); calibrated verdicts on M1–M10; a top-10–15 strongest-sources list with one-line reliability and commercial-position notes; an explicit data-gap list; candidate claims; and the bounded search-stopping rationale. Uncertainty findings are as valuable as positive findings.

Editor’s note (repo workflow): outputs land in raw-inputs/chatgpt-YYYY-MM-DD.md and raw-inputs/gemini-YYYY-MM-DD.md; then verification & critique (critique.md), claim extraction (ledger.md, IDs R07-C001+; next source number S-219), synthesis, and content extraction per PROTOCOL v1.4. Archive audit runs before external research. Corpus anchors: R06-C001 (nutrient systems, not labels — the license), R06-C008 (substrate-dependent optima in coir), R06-C015 (delivery architecture — the keystone R07 extends), R06-C021 (EC is not equivalence), R06-C024/C025 (leaching regime), R06-C027 (flush as discharge), R06-C032 (assay dependence); R08-C006 (the one controlled organic-vs-mineral comparison), R08-C013/C014/C015 (media reuse, adjacent), R08-C017 (pathogens in coir and rockwool), R08-C024 (alkalinity drives pH), R08-C027 (media minor indoor term), R08-C029 (no economic dataset), R08-C030 (no system-level bed-vs-substrate comparison — the absence R07 must not paper over); R04-C009 (ordinance-descended water figure), R04-C018 (indoor water balance), R04-C029 (per-kg reversal), R04-C030 (media as carbon lever), R04-C033 (no media mass-flow inventory), R04 M10 (open); R05-C015 (ET per gram), R05-C024 (metals track media and history), R05-C032/C033 (functional unit, boundary); R01-C018, R01-C033; S-003, S-007, S-096, S-100/S-101, S-104, S-110–S-113, S-116, S-134, S-205. Audit backlog reminder: R03 post-signoff audit remains queued (AUDIT_LOG); at most one backlog audit runs alongside this round.