Claim Ledger — R07: Soil vs. Coco vs. Hydroponic Root-Zone Systems in Cannabis Cultivation

A. Definitions, taxonomy, and physical properties

R07-C002 — Keystone (the standing “system-bundle comparison” example). In the only located controlled comparison of three root-zone systems in drug-type cannabis, an 11-L peat-substrate pot top-fertigated with a synthetic solution (labelled “hydroponic”) produced 42–116% more inflorescence biomass than the other two root-zone systems (both non-synthetic arms) — an 11-L peat pot fertigated with aquaculture effluent (“aquaculture”) and a 3-L peat pot with the lower roots in a deep-water-culture aquaponic sump (“aquaponics”), with significant root-zone effects on most growth and physiological traits. The study cannot support a generic “hydro yields more than soil” claim: the “hydroponic” arm is a peat container, and the three arms differ simultaneously in nutrient source, root-zone geometry and architecture — it is a bundle comparison, not a substrate or architecture comparison.

Evidence status: established (that the bundles differed) · Specificity: direct · Confidence: high (magnitude, abstract-verified) / medium (mechanism attribution impossible by design) · Tag: measured + methodological
Boundary: indoor, southwest Ontario licensed facility · two drug-type cultivars (‘Nordle’, ‘Sensi Star’) · flowering stage · dry inflorescence biomass per plant per cycle (range across cultivars) · no cycle-time or per-m² data · abstract-level (Elsevier paywall; Semantic Scholar abstract)
Descriptor: (1) peat-based substrate, source n/r · (2) 11 L / 11 L / 3 L pots, density n/r · (3) free-draining containers; aquaponic arm roots partly submerged · (4) top fertigation (synthetic) / top fertigation (aquaculture effluent) / DWC aquaponic sump — recirculation status of the sump n/r · (5) n/r · (6) n/r · (7) solution-driven / effluent (biologically mediated) / aquaponic (biologically mediated) · (8) n/r (air-temp-only at best) · (9) water-undisclosed · (10) nutrient program not held constant — unmatched; not root-volume equivalent
Sources: S-219 (Yep, Gale & Zheng 2020b) · Agreement: conflict resolved (ChatGPT CAND-01/02 verified YES; Gemini’s “60 L DWC tubs / ‘Zenith’ / ‘G1’ / 21.5 °C / K-Fe supplementation” descriptor is fabricated — critique §2b-1, do-not-cite)
Notes: Guelph (Zheng) lab. This row is the corpus’s standing illustration of why the eleventh field exists. Full-text figures (per-cultivar yields, cannabinoid results) are owed at v1.0 if the PDF is obtained. Epistemic split (ChatGPT review): the 42–116% inflorescence difference carries the row’s confidence as verified; any explanation of why the synthetic-fertigated peat arm won is lower-confidence because nutrition, root volume, delivery architecture and substrate changed together. Synthesis must not shorthand this as a “peat” or “hydroponic” effect.

R07-C005 — Root-system morphology, architecture and performance in soilless-grown plants respond to environmental and cultivation factors — container volume and root restriction, substrate physical properties, water and oxygen availability, temperature and nutrient supply — as documented in a general horticultural review; this is `general` background for the root-volume and root-zone-oxygen questions and carries no cannabis datum.

Evidence status: established (general) · Specificity: general · Confidence: medium (abstract and review scope verified; not read against a cannabis question) · Tag: methodological (review)
Sources: S-225 (Balliu et al. 2021; identity PASS in verify-A) · Agreement: gemini-only (not used by Gemini for a numbered claim; entered by the editor as background)
Notes: Enters under the R06 precedent for verified Gemini-only reviews. Gemini’s verdict that the container ladder is “obsolete under automated fertigation” is not in the review and is not carried (critique §2b-8). verify-D 2026-09-10: hold lifted — untouched by the correction round; stands on verify-A identity PASS, no Gemini reading carried.

B. Head-to-head evidence in cannabis and hemp

R07-C007 — Substrate response is genotype-specific: in an indoor pot trial of two genotypes across peat, green-fibre and coir, ‘KANADA’ lost more than half its floral yield in coir (3.84 g/plant) and green fibre (4.94 g/plant) relative to peat (8.56 g/plant), while genotype ‘0.2x’ held stable high yields across substrates (9.19 g/plant green fibre to 7.90 g/plant coir). Genotype × substrate interaction is therefore a measured cannabis phenomenon, and a substrate ranking from one genotype does not transfer.

Evidence status: established (interaction exists) · Specificity: direct · Confidence: high (abstract figures verbatim) / medium (generality: two genotypes, one lab) · Tag: measured
Boundary: indoor pot culture, Hohenheim, 36 plants, Latin square · floral yield g/plant per cycle · cannabinoid content also reported (not carried here) · two genotypes (KANADA, 0.2x)
Descriptor: (1) peat mix (PM) / green-fibre 30% blend (G30) / coconut coir (CC), sources per paper · (2) pot volume n/r in abstract · (3) free-draining pots · (4) hand/drip fertigation n/r · (5) n/r · (6) n/r · (7) solution-driven (assumed from design; confirm full text) · (8) air-temp-only · (9) water-undisclosed · (10) nutrient program held constant across substrates (equal program; equivalence basis = equal label rate, unconfirmed)
Sources: S-226 (Burgel, Hartung & Graeff-Hönninger 2020; Crossref abstract) · Agreement: chatgpt-only (CAND-03 YES); Gemini’s data-gap 5 (“multi-genotype trials are absent”) is wrong within its own search (critique §2b-7)
Notes: Together with C014 (two cultivars on stone wool) this is the ledger’s basis for the standing rule that genotype × system is tested, not assumed. Full text owed for pot volume and irrigation regime.

R07-C008 — In a semi-controlled greenhouse comparison of stone wool, peat-perlite and coconut coir under medium-specific fertigation schedules, growing medium altered vegetative growth and assimilate partitioning (leaf-area index and leaf/stem biomass accumulation were impaired in peat-perlite relative to stone wool and coir; the partitioning effect was “marginal but significant”), while the effect of medium on CBD concentration was negligible; the authors attribute the differences mainly to water-holding capacity and irrigation management — i.e., to the bundle, not the material.

Evidence status: probable (single study) · Specificity: direct (chemotype III medical cannabis) · Confidence: medium · Tag: measured — abstract-level (ScienceDirect refused; abstract via mirror, matches publisher snippet)
Boundary: semi-controlled greenhouse, Hohenheim · one chemotype III strain · growing-degree-day growth models · vegetative growth, partitioning, yield formation, CBD concentration · per plant per cycle
Descriptor: (1) stone wool (RW) / peat-perlite (PP) / coconut coir (CC), grades n/r · (2) pot cultivation, volume n/r in abstract · (3) free-draining · (4) drip/hand fertigation n/r · (5) medium-specific schedules — irrigation deliberately not matched · (6) water-holding named as the explanatory variable, values n/r in abstract · (7) solution-driven · (8) air-temp-only · (9) water-undisclosed · (10) nutrient program medium-specific — unmatched by design; system-bundle
Sources: S-220 (Schober et al. 2023) · Agreement: both on the paper (ChatGPT CAND-04 YES; Gemini cited it as “Ahmed et al. 2023” with S-096’s tag — corrected, critique §4-5); Gemini’s “7–14 days faster canopy fill” is not in the paper and is not carried
Notes: The one direct three-class comparison (inert / organic-matter-based (fibrous) soilless × 2) in the corpus, and it is unmatched on irrigation by design — which is the finding. Re-confirm numerics against the publisher PDF before any synthesis figure is quoted (verify-B note 6).

R07-C009 — Across five substrates built from coir, peat and perlite under a common nutrient program, the substrates’ available water was the main factor affecting growth and production; the two coir-fibre + perlite substrates showed the lowest growth. Measured substrate physics — not the substrate’s name — correlated with performance.

Evidence status: probable (single study) · Specificity: direct (hemp-type/chemotype III, var. Tiborszallasi) · Confidence: medium · Tag: measured
Boundary: controlled cultivation, Spain (CEBAS-CSIC) · var. Tiborszallasi · growth and production, CBD yield, aquaporin expression · per plant per cycle
Descriptor: (1) five coir/peat/perlite formulations (Projar-supplied; compositions per paper) · (2) n/r · (3) free-draining containers · (4) n/r · (5) n/r · (6) available water measured per substrate (the explanatory variable) · (7) solution-driven · (8) air-temp-only · (9) water-undisclosed · (10) common nutrient program — held constant
Sources: S-227 (Ortiz-Delvasto, García-Gómez, Carvajal & Bárzana 2023 online / 2024 issue; Crossref abstract) · Agreement: chatgpt-only (CAND-05 YES)
Notes: Supplier-provided substrates; conflict statement to be checked at v1.0. Mechanism candidate for “coco is the best of both worlds”: water availability, which is a grade and irrigation property (C003).

R07-C010 — Within coir, container capacity is a yield variable: in the flowering-stage trial, the drier, lower-container-capacity coir formulation (U2-HP) delivered 29% higher dry floral weight than the higher-container-capacity formulation (U2), and the authors suggest drier coir substrates may be preferred in flowering.

Evidence status: probable (single study) · Specificity: direct · Confidence: medium · Tag: measured — abstract-level (Acta Hortic abstract, actahort.org)
Boundary: walk-in growth chambers, Guelph · medical cannabis, cultivar n/r in abstract · dry floral mass per plant per cycle · liquid organic fertilizer
Descriptor: (1) two commercial coir-based formulations (U2 vs U2-HP, higher perlite fraction) · (2) container volume 6–7.6 L (abstract-level; R06 record says 6 L, Gemini 7.6 L — confirm) · (3) free-draining · (4) hand/drip fertigation, drain-to-waste · (5) irrigation rule n/r in abstract · (6) container capacity measured (U2 > U2-HP) · (7) mineralization-dependent/mixed (liquid organic fertilizer) · (8) air-temp-only · (9) water-undisclosed · (10) same nutrient program across formulations
Sources: S-228 (Caplan, Dixon & Zheng 2019, Acta Hortic 1266:55–62) · Agreement: chatgpt-only (CAND-06 YES)
Notes: Reads with C011 (same lab, same substrates) and C009: water-holding is the recurring correlate across labs — but the direction differs by stage (flowering here favoured drier; C009 favoured higher available water in growth), so no universal “drier is better” is carried.

R07-C011 — Inherits R06-C008 / R08-C007; no patch. Fertilizer-rate optima differed between two coir-based substrates in the flowering trial (212–261 mg N/L, with U2 peaking at the highest rate), a substrate × fertility interaction — while the vegetative trial found no substrate effect and a pooled optimum of 389 mg N/L. Organic-liquid nutrition is therefore demonstrably workable in coir (M8), and its optimum is substrate-specific in flowering.

Evidence status: established · Specificity: direct · Confidence: high · Tag: measured
Boundary: growth chambers, Guelph · ‘Wappa’ · flowering: two coir substrates × five rates (57–283 mg N/L) · floral dry weight per plant per cycle
Descriptor: as C010 (U2 / U2-HP coir; liquid organic fertilizer; drain-to-waste; air-temp-only; water-undisclosed; program varied by design — rate ladder, equal-nominal-N basis)
Sources: S-101 (full-text verified at R08); S-100 for the vegetative trial · Agreement: both on the paper; Gemini’s “substrate-dependent vegetative optimum” and “57–389 mg/L” misread S-100 (critique §4-13/14, §6)
Notes: Critique §5-5: container volume 6–7.6 L abstract-level pending.

R07-C012 — (ledgered from the registry abstract, not from Gemini). In a Thai simplified-media trial with cultivar ‘Platinum Kush Breath Remix’, a peat:perlite:vermiculite 3:1:1 mix significantly enhanced plant height, stem diameter, node number and inflorescence yield, while a coir:perlite:vermiculite:vermicompost 3:1:1:1 mix produced the highest chlorophyll a and total chlorophyll and was comparable to the peat mix in yield. (Gemini reported the opposite on chlorophyll, omitted the vermicompost, and said no yield was measured — all contradicted by the abstract.)

Evidence status: probable (single study; abstract) · Specificity: direct · Confidence: low (abstract-only; PDF text not extractable; n, pot volume and durations unverified) · Tag: measured — abstract-level
Boundary: Thailand, controlled cultivation · one cultivar · height, stem diameter, nodes, chlorophyll, inflorescence yield · per plant per cycle; time axis n/r
Descriptor: (1) peat:perlite:vermiculite 3:1:1 vs coir:perlite:vermiculite:vermicompost 3:1:1:1 (and other mixes per paper) · (2) n/r (Gemini’s “12 L air-flow pots” unverified) · (3) free-draining · (4) n/r · (5) n/r · (6) n/r · (7) mixed (vermicompost arm partly mineralization-dependent) · (8) air-temp-only · (9) water-undisclosed · (10) n/r
Sources: S-229 (Tebdoie et al. 2025, IJAT 21(4):1521–1534, DOI 10.63369/ijat.2025.21.4.1521-1534 — identity PASS, verify-A) · Agreement: gemini-only (claim support inverted; editor re-read)
Notes: Enters as a real cannabis substrate comparison the ChatGPT pass did not find. Stays abstract-level until the full text is read; no synthesis figure is drawn from it before then. verify-D 2026-09-10: hold lifted — Gemini item 8 now reproduces the registry abstract exactly; row remains abstract-level, confidence cap unchanged.

R07-C013 — Keystone (extends R06-C015, confirmed in full text). Delivery architecture is an independent variable at fixed medium and nutrition: in medicinal cannabis fertigated with a standard or P/K/Fe-augmented solution (+83% P, +39% K, +860% Fe), neither the augmented regime nor the fertigation system had a statistically conclusive effect on final inflorescence harvest mass, but the recirculating system consistently outperformed drain-to-waste on total THC yield, especially after the ninth week, and the authors propose week 11 as the harvest time for recirculation. No universal recirculation yield advantage is established; the effect is on total cannabinoid yield and harvest timing.

Evidence status: probable (single study, sequential cycles) · Specificity: direct · Confidence: medium · Tag: measured
Boundary: indoor, Prague · medicinal cannabis (THC-dominant) · 27.5 plants/m² · inflorescence mass per plant per cycle; total THC yield per plant; 7-day vegetative phase; generative phase sampled weekly through week 13; fertigation through week 9, demineralized water from week 10; harvests to week 13 · the two architectures were run as sequential cycles, not side by side
Descriptor: (1) 3.0 L expanded clay (LECA) in 3.45 L pots — inert-substrate fertigation class · (2) 27.5 plants/m² · (3) drain-to-waste vs recirculating · (4) drip fertigation; recirculating vs DTW is the treatment · (5) per paper · (6) n/r · (7) solution-driven · (8) air-temp-only (air 25/22 °C; root-zone temperature not measured) · (9) demineralized water base (partially disclosed) · (10) nutrient program held constant across architectures within each regime; augmentation basis = equal base solution + additions
Sources: S-205 (full text PMC10847352; verify-B CAND-24 YES; verify-A full-text re-read) · Agreement: both on the paper — Gemini’s “+182% vs baseline,” “+93/+43/+940%,” “0.5 m²/plant” corrected to the paper’s comparators (critique §2b-2)
Notes: R06-C015’s wording (“total THC yield,” not “yield”) stands. The 182% figure is recirculating-vs-DTW; 50.7% is augmented-vs-control within DTW at week 11 only. Sequential design means season/cycle is confounded with architecture — a rule-12-adjacent caution for synthesis. verify-D: Gemini’s “11-week flowering cycle” not adopted (do-not-cite string).

R07-C014 — Crop steering by osmotic stress on stone wool reduced plant height by 15% in both cultivars under high-EC and hybrid treatments, and cut flower yield by ~20% in a cultivar-dependent pattern (‘Cherry’ under both high-EC treatments; ‘Trump’ only under the hybrid 8/4 mS/cm treatment), with no difference in cannabinoid concentration in either cultivar — no compensating potency gain. (CAND-09 restated per critique §3 B-09: height effect uniform, yield effect cultivar-dependent.)

Evidence status: probable (single study) · Specificity: direct · Confidence: medium-high (full text) · Tag: measured
Boundary: controlled environment, Utah State · medical cannabis, two cultivars · height; flower yield per plant per cycle; cannabinoid concentration · EC treatments 8/4 and 8 mS/cm vs control
Descriptor: (1) stone wool, Grodan · (2) 15 × 15 cm blocks · (3) free-draining, drain-to-waste · (4) drip fertigation · (5) per treatment schedule · (6) n/r · (7) solution-driven · (8) air-temp-only · (9) water-undisclosed · (10) program varied by design — EC ladder at fixed composition
Sources: S-230 (Allred, Fatzinger & Bugbee 2025; full text Europe PMC) · Agreement: chatgpt-only (CAND-09 PARTIAL → corrected)
Notes: Bugbee lab (with S-104). Second measured genotype × system interaction (with C007). “Crop steering” claims (G22) enter the ledger with this row only.

R07-C015 — Cannabis drought and crop-steering responses are not directionally uniform: controlled drought raised THCA and CBDA concentration by 12–13% and THCA yield by 43% in one Guelph study (S-231); high osmotic stress on stone wool reduced yield without a concentration gain (S-230, C014); moderate deficit irrigation cut water use without biomass or CBD-yield loss while severe stress impaired performance (S-182, C016). The corroboration is of heterogeneity — three studies, three genotype sets, three root-zone bundles — not a pooled effect size.

Evidence status: established (that responses diverge) · Specificity: direct · Confidence: high · Tag: measured (synthesis of three)
Boundary: three different systems (coir / stone wool / indoor pots), genotypes and stress definitions; not poolable by construction
Sources: S-231 (Caplan, Dixon & Zheng 2019, HortScience 54(5); Crossref abstract), S-230, S-182 · Agreement: chatgpt-only (CAND-08 YES)
Notes: R11 handoff: water-status responses belong to the environment-response report; R07 carries only the system-dependence point.

R07-C016 — Moderate drought stress reduced water consumption without negatively affecting biomass or CBD yield in two chemotype III genotypes grown indoors, while severe drought consistently impaired plant performance and provided no agronomic advantage. This is an irrigation-management water saving at fixed medium — not a medium effect (M5).

Evidence status: probable (single study) · Specificity: direct (chemotype III) · Confidence: medium · Tag: measured
Boundary: indoor, Hohenheim · genotypes Kanada, Terra Italia · water consumption per plant per cycle; biomass; CBD yield · abstract-level (Frontiers page + abstract; full text open — numerics to be added at v1.0)
Descriptor: (1) substrate per paper (organic-matter-based (fibrous) soilless class; grade n/r) · (2) n/r · (3) free-draining · (4) n/r · (5) deficit-irrigation regimes are the treatment · (6) n/r · (7) solution-driven · (8) air-temp-only · (9) water-undisclosed · (10) nutrient program held constant
Sources: S-182 (Crispim Massuela et al. 2026) · Agreement: chatgpt-only (CAND-10 YES)
Notes: Different paper from S-099 (Massuela 2023) — do not conflate.

R07-C017 — Subsurface drip irrigation reduced irrigation water use by 18.6% relative to surface drip (188.19 vs 231.23 L/m²) in an outdoor foil-tunnel experiment with three CBD-rich chemotype III genotypes, with inflorescence yield up 5% and CBD concentration up 9%. A delivery-architecture effect on water use — outdoor, in-ground, not transferable to indoor water culture.

Evidence status: probable (single study) · Specificity: direct (chemotype III, outdoor) · Confidence: medium-high (full text) · Tag: measured
Boundary: outdoor foil tunnel, Hohenheim · Kanada, Terra Italia, FED · water applied per m² per season; inflorescence yield per m²; CBD concentration
Descriptor: (1) field soil in tunnel (soil class) · (2) in-ground, density per paper · (3) in-ground, drainage n/r · (4) subsurface vs surface drip; drain-to-waste (no recapture) · (5) per paper · (6) n/r · (7) per paper · (8) air-temp-only · (9) water-undisclosed · (10) nutrient program held constant across architectures
Sources: S-180 (Büser, Hartung & Graeff-Hönninger 2025; full text Europe PMC) · Agreement: chatgpt-only (CAND-11 YES)
Notes: The only measured cannabis water-applied comparison by architecture located; the metric is water applied per m² per season, not water withdrawn per kg flower (R04 framing).

C. Mechanisms in the root zone

R07-C019 — (ledgered from the abstract, not from Gemini). Under NaCl stress, drug-type cannabis growth and physiology in a hydroponic nutrient solution — but not in the aquaponic solution — became negatively affected at 40 mM NaCl, while cannabinoid concentrations declined linearly with NaCl in both solutions. Solution chemistry (biologically mediated vs mineral) modulated salt stress; Gemini reported the direction inverted.

Evidence status: probable (single study) · Specificity: direct · Confidence: medium (open-access full text exists; ledgered from abstract pending full read) · Tag: measured — abstract-level
Boundary: Guelph, controlled environment · drug-type cannabis · NaCl ladder to 40 mM · growth, physiology, cannabinoid concentration · per plant; time axis n/r
Descriptor: (1) water-culture / solution comparison (substrate per paper) · (2) n/r · (3) n/a · (4) hydroponic vs aquaponic solution · (5) n/r · (6) n/a · (7) solution-driven vs biologically mediated · (8) n/r · (9) NaCl-adjusted solutions (disclosed by treatment) · (10) base nutrition differs by design
Sources: S-232 (Yep, Gale & Zheng 2020a; PMC7424260; identity PASS both verifiers) · Agreement: both cite the paper (ChatGPT R12 for CAND-23 feasibility only; Gemini A3 inverted) — the salinity reading here is verify-A’s
Notes: verify-D 2026-09-10: hold lifted — Gemini item 9 corrected the salinity direction to the abstract’s (hydroponic, not aquaponic, solution affected at 40 mM); row unchanged, still abstract-level pending full read.
Notes: Bears on coir Na/Cl load (C003/C004) as an adjacent-within-cannabis salt-stress datum. Full-text read owed at verify-D.

R07-C021 — (structural, no numeric cells). The R07 synthesis target is a system-property matrix, not a winner: root-zone systems are mapped on 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 · media longevity and reuse · monitoring burden · failure velocity · automation compatibility · capital and labor intensity — with each cell graded `direct` / `adjacent` / `general` / `unmeasured`. Gemini’s contribution retained here is the scaffold only: the A1 taxonomy layout and the H24 matrix structure, plus the observation that time-to-symptom and time-to-remediate are different axes (a forgiving medium may show a problem late and also fix it slowly). Every numeric cell Gemini placed in H24/H25/H26 (yields per m² per cycle and per year, cycles per year, kg media per plant, L per kg flower, embodied-carbon ranges, failure hours) is unsourced and not ledgered; in this matrix those cells read `unmeasured`.

Evidence status: n/a (structural) · Specificity: n/a · Confidence: n/a · Tag: methodological
Sources: brief v1.0 §H24 (primary); Gemini A1 / H24 as scaffolding (raw-inputs/gemini-2026-09-09.md; verify-A §4-13/14 for the withdrawn cells) · Agreement: both adopt the matrix (ChatGPT H24 section; Gemini H24) — cells sourced from this ledger only
Notes: Enters the ledger so the synthesis can cite a structure rather than re-derive it; held at v0.1 because its only Gemini-specific content (two-axis failure framing) awaited the correction round. verify-D 2026-09-10: hold lifted — Gemini item 24 declares every H25/H26/F2/F5/G1 table range illustrative and unsourced; cells stay unmeasured. The time-to-symptom / time-to-remediate axis is unmeasured in cannabis (C020). The matrix’s analytic axes (buffering, root-zone oxygen control, nutrient-supply precision, water-use control, recirculation capability, biological dependence, media longevity/reuse, monitoring burden, failure velocity incl. time-to-symptom vs time-to-remediate, automation compatibility, capital/labor intensity) are editorial structure; each cannabis cell is graded separately and reads unmeasured unless a ledger row fills it. Matrix completeness is not evidence completeness (ChatGPT review item 7).

D. Quality, chemistry, and sensory

R07-C022 — Keystone (the strongest R07 absence; M2 verdict). No controlled root-zone study using a trained sensory panel (class b) or a blinded consumer-preference test (class c) — soil-grown vs soilless vs water-culture flower, genotype and environment controlled — was located within the documented search (through 2026-09-09). “Soil tastes better,” “hydro is bland” and “coco gives cleaner terps” are three hypotheses with class-(d) evidence only; the archive never made the “soil tastes better” claim at all.

Evidence status: unknown · Specificity: direct (absence) · Confidence: high · Tag: absence
Sources: both searches (ChatGPT CAND-15, gap 3; Gemini C013, re-worded from “zero published … globally” to the bounded form); archive-audit (M2 absent from the archive) · Agreement: both (critique §2a-1)
Notes: R06-C022 is the fertility twin. Gemini’s “Vergara et al. 2016” sensory citation is unresolved and not carried.

R07-C023 — Chemistry differences across substrate or root-zone treatments (class a) do not by themselves establish flavor or preference differences (classes b/c): the cannabis papers that report cannabinoid or terpene concentration by substrate — S-219 (three bundles), S-226 (two genotypes × three substrates), S-220 (CBD insensitive to medium), and a substrate × microbial-biostimulant factorial (S-233) — report chemistry endpoints only and no sensory outcome.

Evidence status: established · Specificity: direct · Confidence: high · Tag: methodological
Sources: S-219, S-226, S-220, S-233 (Lyu et al. 2025; abstract) · Agreement: chatgpt-only (CAND-16 YES); consistent with R06-C023
Notes: R12 handoff (which compounds move with substrate: THCA/CBDA and terpene concentrations are reported in S-219/S-226/S-233 and are to be tabulated there, not here).

E. Plant health, pathogens, and safety

R07-C024 — Keystone (extends R08-C017; two independent labs). Soilless media are not pathogen-free in production: pathogenic Pythium, Globisporangium and Fusarium isolates were recovered from coconut coir and stone wool where cannabis with root-rot and wilt symptoms was grown in a commercial Connecticut facility, and all isolates tested were pathogenic to hemp seedlings in vitro and in vivo (disease incidence 6.7–100%); independently, several Pythium species were shown to cause crown and root rot on cannabis under commercial greenhouse conditions in British Columbia. M9’s premise (“soilless media are sterile, so fewer root diseases”) is rejected; the opposite (“soil has fewer diseases”) is also unsupported because no controlled incidence ranking by medium on a common denominator was located within the documented search (cutoff 2026-09-09).

Evidence status: established · Specificity: direct · Confidence: high · Tag: measured (+ absence for the ranking)
Boundary: commercial cannabis facilities (Connecticut; British Columbia) · pathogen recovery, Koch’s postulates on hemp seedlings · no incidence rate by medium
Descriptor: (1) coconut coir and stone wool (S-116); greenhouse substrate per paper (S-234) · (4) architecture per facility, n/r · (8) n/r · (9) water-undisclosed · (10) n/a — surveys, not comparisons
Sources: S-116 (McGehee & Raudales 2021, full text), S-234 (Punja, Scott & Lung 2021 online / 2022 print, CJPP 44(1):66–81; abstract) · Agreement: both (ChatGPT CAND-12; Gemini E1 — but Gemini credited S-116’s species list to Punja and gave Punja a fabricated locator; corrected, critique §4-4, §4-22)
Notes: Raudales lab (UConn) + Punja lab (SFU) = rule 12 satisfied for the recovery finding. Detection methods: culture and morphological/molecular identification in both. R20 (IPM) handoff carries the archive’s “soil spreads bugs” claim, which has the same absent denominator.

R07-C025 — Hydroponically grown cannabis sampled over three years (2014–2017) developed root rot caused by Pythium dissotocum and P. myriotylum and Fusarium oxysporum infection, and the potential for F. oxysporum spread through the hydroponic system was confirmed by its detection in the recirculating nutrient solution — recirculation is a documented inoculum route, not only a hypothesised one.

Evidence status: established · Specificity: direct · Confidence: high (abstract-verified) · Tag: measured — abstract-level (T&F paywall)
Boundary: licensed commercial hydroponic cannabis, British Columbia, 2014–2017 · root-disease etiology and solution detection · no incidence rate by architecture
Descriptor: (1) substrate per facility (inert-substrate fertigation / water culture — meaning (b)/(c); re-file at full text) · (4) recirculating · (8) solution temperature and DO n/r in abstract · (9) water-undisclosed
Sources: S-235 (Punja & Rodriguez 2018, CJPP 40(4):498–513) · Agreement: both (ChatGPT CAND-13; Gemini E1)
Notes: The temperature × dissolved-oxygen × oxygen-demand × pathogen interaction (brief C8) remains unmeasured in cannabis in every located source; it is adjacent CEA hydroponics and practitioner experience (C035). Reservoir-temperature thresholds are convention (chillers “in the more professional rooms,” answer 2b).

R07-C026 — Stone-wool fibre carries occupational exposure limits: NIOSH REL TWA 3 fibers/cm³ (fibres ≤3.5 µm diameter, ≥10 µm length) and 5 mg/m³ total; OSHA PEL-TWA as nuisance dust 15 mg/m³ total / 5 mg/m³ respirable; ACGIH TLV 1 f/cc for rock wool; rock (stone) wool is “not classifiable as to carcinogenicity to humans” on OSHA’s health-effects page. No cannabis-specific worker-exposure measurement for stone wool, peat dust or reservoir biofilm was located within the documented search (cutoff 2026-09-09).

Evidence status: established (limits) / unknown (cannabis exposure) · Specificity: general (regulatory) · Confidence: high · Tag: regulatory (U.S. occupational guidance and limits — NIOSH and OSHA values; ACGIH TLV, a non-governmental body — as of 2026-09-10) + absence
Sources: S-236 (NIOSH Pocket Guide, mineral wool fiber), S-237 (OSHA Synthetic Mineral Fibers pages) — both fetched live · Agreement: chatgpt-only (E15)
Notes: Jurisdiction- and date-stamped; state (Cal/OSHA) equivalents not checked. R20/R21 handoff. Rule-13 wording correction 2026-09-11 at synthesis review (ChatGPT catch, editor-confirmed); substance unchanged.

R07-C027 — Inherits R06-C004. As of 2026-09-09, California’s OCal program certifies cannabis to comparable-to-organic standards (CDFA cultivation; CDPH manufacturing), and cannabis cannot carry the USDA “organic” designation because it is federally illegal and ineligible under the NOP; the word “organic” cannot be used for cannabis grown or manufactured in California outside OCal.

Evidence status: established · Specificity: direct · Confidence: high · Tag: regulatory (California, as of 2026-09-09)
Sources: S-215 (CDFA + CDPH, re-fetched 2026-09-10) · Agreement: both (ChatGPT CAND-22 YES; Gemini E2 for DCC §15723 metals — correct section, not ledgered separately here; R05/R08 carry contaminant limits)
Notes: Certification is separate from technical feasibility (C028).

F. Water, nutrient, waste and footprint flows

Evidence status: probable · Specificity: adjacent (horticultural media LCAs; hemp LCA) · Confidence: medium · Tag: modeled
Boundary: per m³ of medium (S-238, S-240) — not per kg flower; scenario-specific system boundaries; Spain / Europe / Latvia; no cannabis LCA varies medium at fixed environment (R04-C030 / R08-C027 media-term reversal inherited)
Sources: S-238 (Toboso-Chavero et al. 2021), S-239 (Paoli et al. 2022), S-240 (Alejandre et al. 2026), S-185 (Kalkušová et al. 2025) — S-238/S-240 abstract-level (Springer/Elsevier), S-239 title-level · Agreement: chatgpt-only (CAND-19 YES with qualifier); Gemini’s F5 embodied-carbon table is unsourced and not carried
Notes: M10 verdict: not supportable universally; per-area vs per-product reversal (R01-C018, R04-C029) applies.

G. Practice, skill, labor and economics

Evidence status: established (as reported) · Specificity: adjacent (economics; root zone not varied) · Confidence: medium · Tag: observed-administrative (facility economics model)
Boundary: New York greenhouse · returns per ft² per year · cycles per year by flowering strategy · one facility model; root-zone system single and constant (descriptor: not applicable — single system; substrate/architecture per paper)
Sources: S-011 (full text Europe PMC) · Agreement: chatgpt-only (CAND-21 YES)
Notes: The only cycles-per-year datum in the corpus, and it is a strategy variable.

R07-C034 — Keystone (re-confirms R08-C030 in a second independent search). No controlled cannabis comparison of persistent living-soil beds (or any soil/amended-bed system) against a conventional soilless or water-culture system across multiple cycles — with genotype, environment and nutrition matched — was located within the documented search (cutoff 2026-09-09); every “soil vs hydro” outcome in the corpus is a single-cycle container comparison (C002, C007, C008, C009, C012) or a management comparison at fixed medium (C013–C017).

Evidence status: unknown · Specificity: direct (absence) · Confidence: high (that none was located; two searches) · Tag: absence
Sources: ChatGPT gap 8; Gemini C014 (re-worded as bounded); R08-C030 · Agreement: both (critique §2a-5) — inherited anchor, independently re-searched
Notes: R13–R16 (soil ecology) handoff. The “soil” class in the R07 literature is almost entirely bagged peat-based mixes (soilless by physical definition) or field soil under tunnel (C017); native-soil or amended-bed indoor systems have no located trial.

Pre-registered public claims — ledger verdicts (M1–M10)

Claim	Verdict	Claims
M1 Hydro grows faster and yields more than soil	Not supportable as a category claim — no cannabis cycle-time or cycles-per-year datum by root-zone system; the one three-system yield comparison is a bundle (peat-container “hydroponic” arm); the archive’s 4–5 vs 1–2 was indoor-vs-outdoor	C018, C002, C001, C033
M2 Soil-grown tastes better / has more terpenes	Unsupported — no class-(b) or (c) study; class-(a) chemistry differs by bundle but does not carry preference; archive never made the claim	C022, C023
M3 Coco is the best of both worlds	Not evidence-based as “best” — coir grade/source, water availability and container capacity produce measured advantages and penalties by genotype and stage; the property behind the slogan is water availability/buffering, not the name	C003, C004, C007, C009, C010
M4 Hydro is unforgiving; soil is easiest to learn	Unverified — plausible buffering hypothesis; no cannabis failure-velocity dataset; practitioner direction only	C020, C021
M5 Hydroponics uses 90% less water	Unsupported as a cannabis statistic — leafy-greens figure; cannabis savings located are irrigation-management effects (−18.6% SDI; deficit irrigation), not medium effects	C029, C016, C017
M6 Hydro leads to a lot of waste — rockwool to landfill	Direction plausible, magnitude unmeasured — no mass-flow inventory; Gemini’s kg/plant not sourced	C030
M7 Soil buffers mistakes; hydro shows problems in hours	Unverified in cannabis — general substrate physics supports buffering differences; time-to-symptom and time-to-remediate are separate, unmeasured axes	C020, C021, C004
M8 You can’t grow organic in hydro/coco	False as an absolute — organic-liquid dose-response in coir; aquaponic cannabis to harvest; certification (OCal) separate	C028, C011, C027
M9 Soilless media are sterile, so fewer root diseases	Premise rejected — pathogens recovered from coir, stone wool and recirculating solution (two labs); opposite also unsupported (no incidence ranking by medium)	C024, C025
M10 Living soil has a lower footprint than coco/hydro	Not supportable universally — media LCAs reverse by indicator, geography and functional unit; no cannabis LCA varies medium at fixed environment; inherited R04/R06 M10	C031, C030, C029

Retired / not carried from raw inputs

Gemini R07-C008 (Yep “60 L DWC / ‘Zenith’ / ‘G1’ / 21.5 °C / K-Fe supplementation”) — withdrawn; fabricated descriptor (do-not-cite). — Gemini C004 Velechovský magnitudes (+182% vs baseline; +93/+43/+940%; 0.5 m²/plant) — replaced by full-text comparators (C013). — Gemini C001 “7–14 days faster canopy fill,” C002 “flowering duration genetically invariant,” C016 “4.8–5.4 vs 3.5–4.2 cycles/yr” — unsourced, withdrawn (C018). — Gemini C005 “CEC 40–100 cmol_c/kg” and “mandatory calcium buffering” — number unsourced, rule not carried (C004). — Gemini C010 “0.30–0.50 kg rockwool/plant, >90% landfill” — unsourced (C030). — Gemini C011 runoff N/P loads and C012 “70–80% / 60–75%” water figures — not in the cited sources (C029). — Gemini C015 “30–50% TYMC failure for living soil” — cited to a Pythium etiology paper; withdrawn. — Gemini C006/C007 Caplan 2017a/b readings — corrected to the abstracts (C011). — Gemini H25 / H26 / F2 / F5 / G1 numeric tables — illustrative, unsourced; not ledgered (C021). — Gemini “Caplan, Stemeroff, Dixon & Zheng 2019 HortTechnology,” “Knight et al. 2010 …04.015,” “Vergara et al. 2016,” “Zheng (2021)” — identity FAIL / unresolved; nothing carried. — ChatGPT CAND-09 “cultivar-dependent height reduction” — restated (C014). — ChatGPT CAND-19 “including rockwool” — qualified abstract-level (C031). — ChatGPT E15 fibre limits carried (C026); no other ChatGPT item retired.

Source file in the research repository: research/reports/R07-soil-vs-coco-hydroponic/ledger.md. Private links, some names and practitioner-record passages are removed from the public copy, and broken character encoding from file conversion is repaired; nothing else is changed.