Research Brief — R05: Indoor vs. Greenhouse/Light-Dep vs. Sun-Grown Cannabis

Report ID: R05 (pilot report) · Brief version: v1.1 — approved August 25, 2026 (v1.0 + five methodology additions from ChatGPT’s editorial review) Version note for critique stage: Gemini’s run launched on v1.0 (before these additions); ChatGPT runs v1.1. Account for this asymmetry when comparing outputs — v1.1 asks for greenhouse subtypes, functional units, lifecycle boundaries, and candidate claims that Gemini was not prompted for. Date: August 25, 2026 · Drafted by: Claude (editor) Assigned to: ChatGPT deep research + Gemini deep research (independently, same brief)

Assignment

Produce an evidence review comparing the three principal cannabis cultivation environments — indoor (sealed/warehouse), greenhouse (including light-deprivation), and outdoor sun-grown — across environmental footprint, product quality, yield and economics, and cultivation risk. This is not a “which is best” piece: the goal is to establish what the evidence actually supports, where it is thin, and where popular belief outruns the data.

This report is the pilot for a larger research program on organic and sustainable cannabis cultivation, so favor depth and citation quality over breadth. Target length: whatever the evidence supports — expected 20–50 pages.

Methodology requirements

Methodological hierarchy. Prioritize direct comparative cannabis studies where available. Where direct cannabis comparisons do not exist, use cannabis-specific observational/LCA evidence next, then adjacent controlled-environment horticulture, then general plant science. Do not imply that indirect evidence proves a cannabis-specific outcome. Clearly label modeled results, observational datasets, controlled experiments, and practitioner claims separately — they are different kinds of evidence and must remain distinguishable in your report.

Functional unit consistency. Whenever comparing environmental or production metrics, identify the functional unit used by each source (per kg dried flower, per m², per plant, per gram THC, per dollar revenue). Prefer normalized comparisons per kg of saleable dried flower where available, while preserving other units where conversion would require unsupported assumptions. Never merge figures across functional units without saying so.

Carbon accounting boundaries. For every energy/carbon estimate, note the study’s lifecycle boundary: operational electricity/fuels only, or also embodied impacts (structures, HVAC and lighting equipment, fertilizer production, growing media, water infrastructure, packaging, transport). Two estimates can disagree dramatically simply because they answer different lifecycle questions — surface the boundary, don’t average across it.

Key questions

A. Definitions & landscape

1. Define the three systems precisely. For greenhouses, do NOT treat “greenhouse/light-dep” as one system: distinguish, wherever the evidence permits, (a) passive greenhouse / hoop house, (b) light-deprivation greenhouse, (c) greenhouse with supplemental lighting, and (d) highly controlled sealed greenhouse. A supplemental-light greenhouse running HVAC and LEDs can look environmentally very different from a passive hoop house; report them separately and let the editor collapse categories later if the evidence demands it. Also explain what “light deprivation” actually involves mechanically and photobiologically.
2. What share of licensed North American production does each system represent, and how has that shifted? Include any solid data on unlicensed/legacy production if available.

B. Environmental footprint

3. Energy: What do lifecycle assessments (LCAs) and utility data show for energy use and greenhouse-gas emissions per unit of dried flower across the three systems? Anchor on peer-reviewed LCAs (e.g., the 2021 Nature Sustainability indoor-cannabis carbon study and anything newer, 2021–2026) — report ranges, assumptions, and geographic sensitivity, not single headline numbers.
4. How much do modern LED fixtures, climate, and grid carbon-intensity change the indoor picture vs. older HPS-era studies?
5. Water: comparative water use per unit yield, including irrigation efficiency, runoff/discharge issues, and regional context (e.g., drought-stressed regions).
6. Land use, ecosystem impacts (especially of outdoor expansion into sensitive areas), pesticide drift, light pollution (greenhouses), and waste streams (media, plastics).

C. Product quality

Treat “quality” as three distinguishable things: chemical quality (cannabinoid/terpene content and diversity), physical/commercial quality (flower morphology, density, appearance, aroma retention, market-grade acceptance), and consumer perception (what buyers believe and pay for). Do not let “indoor is higher quality” quietly reduce to THC percentage.

7. What does controlled or credible comparative evidence say about cannabinoid potency across systems? Distinguish measured differences from market perception.
8. Same for terpene/secondary-metabolite profiles, including UV exposure and environmental-stress hypotheses. Is “sun-grown has richer terpene profiles” or “indoor is more potent” actually supported by direct study?
9. Contamination risk by system: mold/mildew, pesticides, heavy metals — what do testing-lab datasets and studies show?

D. Yield, economics & risk

10. Comparative yield per plant / per square meter / per cycle, and cycles per year (including light-dep multi-harvest).
11. Production cost per unit across systems; capital vs. operating cost structure; how falling wholesale prices interact with each model’s viability. Treat economic comparisons as highly jurisdiction- and time-sensitive: a cost estimate from California, Canada, Colorado, or Massachusetts may be dominated by electricity price, labor, taxes, licensing, required testing, building codes, and wholesale market conditions rather than cultivation biology. Separate cultivation-system economics from regulatory, labor, energy-price, tax, and market-price effects whenever possible.
12. Cultivation risk profile of each system: pest and disease pressure, weather exposure, crop failure modes, climate-change sensitivity.

E. Synthesis targets

13. Where do the systems genuinely differ on evidence, where are differences negligible, and where is the honest answer “not studied”?
14. What are the 5–10 most repeated public claims on this topic (“indoor is stronger,” “outdoor is automatically greener,” “light-dep is the sustainability sweet spot”), and what does the evidence say about each?

Known controversies to handle carefully

• Indoor energy figures are frequently quoted from pre-LED, pre-legalization studies — check vintage and assumptions of every number.
• “Sun-grown = sustainable” ignores water/land/ecosystem impacts in some regions; don’t let a green halo substitute for data.
• Potency comparisons are confounded by genetics, harvest timing, and lab variability — flag confounds explicitly.
• Quality claims are heavily marketed by whichever segment benefits; watch for industry-funded sources and label them.

Source & citation requirements

• Prefer: peer-reviewed studies and LCAs, university extension and government reports (state cannabis agencies, DOE/utility studies), testing-lab datasets. Industry white papers and journalism are acceptable for market data if labeled as such.
• Provide full citations (authors, year, title, journal/publisher, DOI or URL) for every substantive claim. No orphan statistics.
• For each major claim, note whether the evidence is cannabis-direct, adjacent-crop/CEA, general plant science, or anecdotal — this distinction is preserved downstream.
• Report ranges and disagreements between sources, not averaged-away conclusions. If two credible sources conflict, present both.
• Explicitly list data gaps — questions this report could not answer from available evidence.
• Note publication year sensitivity: the industry changes fast; date-stamp claims where it matters (esp. energy and market share).

Candidate claims for editorial review

Conclude with 15–30 concise, evidence-based claims that appear sufficiently important and reusable to merit possible inclusion in the R05 Claim Ledger, each phrased as a testable statement with its supporting citation(s). Do not assign final evidence status or confidence; the editor will do that.

What “done” looks like

A structured report answering questions A–E with cited evidence, a stated-limitations section, a list of the strongest sources (top 10–15 with one-line reliability notes), an explicit data-gap list, and the candidate-claims section above. Do not smooth over uncertainty — uncertainty findings are as valuable as positive findings.

Editor’s note (repo workflow): outputs go to raw-inputs/chatgpt-YYYY-MM-DD.md and raw-inputs/gemini-YYYY-MM-DD.md. Claude then runs citation spot-checks, cross-examination (critique.md), claim extraction (ledger.md, IDs R05-C001+), synthesis, and content extraction per PROTOCOL.md.