R02 — The Global Cannabis Cultivation Landscape

CCKE RESEARCH REPORT · R02 — Where cannabis is grown, under which systems, at what scale — and what the evidence can actually support

Frozen Brief v1.0 · PROTOCOL v1.3 · ChatGPT independent deep-research pass · Research cutoff: 31 August 2026

(Local mirror of the Drive original, chatgpt-2026-08-31.docx; text captured verbatim 2026-08-31. The .docx in Drive raw-inputs/ is canonical for formatting. Full provenance-record tables preserved in the docx; key fields summarized here where tables were structural.)

Reading note

This report is deliberately strict about denominators. It separates licensed from unlicensed production, hemp from drug-type cannabis, canopy from output, wet mass from dry flower, and detected activity from total cultivation. Policy effects and enforcement consequences are reserved for R03.

Executive finding

Cannabis is cultivated almost everywhere, but it is not measured as one global crop. UNODC’s 2026 synthesis reports evidence of cultivation in at least 155 countries, covering about 96% of the world’s population. That statement describes geographic reach, not hectares or tonnes. No current source provides a defensible worldwide total for drug-type cannabis area or dry-flower production. The visible record is a mosaic: audited or mandatory administrative reporting in a few regulated markets; crop surveys in a small number of traditional producing regions; seizures and eradication records elsewhere; and large blank areas where cultivation is inferred from use, trafficking, or reputation.

The practical conclusion is not that the landscape is unknowable. It is that different questions have different evidence. Canada can report licensed indoor and outdoor area monthly. California and Oregon can report track-and-trace harvests and wholesale transfers, though data-entry errors and inactive licenses complicate interpretation. Morocco’s Rif and Afghanistan have survey-based historical anchors, but their current output cannot be ranked reliably. EUDA can characterize broad supply patterns, while warning that seizures and stated origin do not equal production. Home cultivation is measurable through registries in Uruguay and surveys in Canada, yet largely invisible elsewhere. Traditional “landrace” regions are historically important, but commercial names are not genetic certificates.

The report’s central rule: never compare numbers until their population, product, period, unit, and estimation method match. A hectare of outdoor licensed canopy, a seized plant, a tonne of resin, a retail unit, and a modeled illicit-market value are not interchangeable evidence.

Methods and boundaries

Prioritized national/subnational administrative datasets, UNODC and EUDA publications, and peer-reviewed geography, genomics, and ethnobotany. Industry reports and journalism used only to trace public claims. Major quantitative estimates receive the ten-field provenance record. “Drug-type/regulated cannabis” includes THC-dominant and other cannabinoid-rich floral production governed as cannabis in the reporting jurisdiction; hemp figures excluded unless a source cannot separate them. Licensed area ≠ harvested area ≠ production mass; wet mass, dry flower, resin, extract, plants, canopy, licenses, sales units, and value kept separate. Confidence labels describe fitness for the stated claim, not institutional prestige.

A. The map

A global crop without a global denominator

UNODC’s current evidence supports ubiquity (most geographically widespread illicitly cultivated drug crop), not a current global tonnage. The 2008-era global range — ~200,000–641,800 ha; 13,300–66,100 t herb; 2,200–9,900 t resin — combined member-state reports, seizures, prevalence, and yield assumptions; useful as a history of estimation, not a contemporary baseline. Provenance record A1 (155 countries / ~96% population; WDR 2026; high for presence, no inference to hectares) and A2 (legacy range; low; historical orientation only) as tabled in the docx.

Regional evidence grades

North America — high for regulated slices, low for total cultivation. Canada (Dec 2024): licensees reported 1,453,902 m² indoor growing area and 596 ha outdoors — areas where cultivation occurred at licensed sites, not national harvested-area or yield. US: no comparable national series; CA and OR unusually useful state data covering licensed activity only.

Latin America & Caribbean — mixed, generally weak for total output. Paraguay, Colombia, Mexico, Jamaica recur in seizure/origin/eradication and qualitative supply reporting; “reported source” is not a production census. Colombia’s licensed medical/export sector administratively visible; unlicensed cultivation not measured comparably to coca monitoring. Mexico’s estimates entangled with eradication/enforcement records.

North Africa — moderate historical anchors, low current total. Morocco 2003 survey: 134,000 ha, 47,000 t raw cannabis, 3,080 t potential resin (Provenance A3; moderate historical, low present-day). Later official estimates: ~47,000 ha for much of 2010–2018; 21,000 ha in 2019; resin ~700–760 t (2010–2017), 424 t (2018), 596 t (2019). Seizures in Morocco/Algeria/Spain approached or exceeded modeled output in some years; hybrid replacement made older kif-based yield factors obsolete. Current rank and tonnage uncertain though Morocco remains a major resin source for Europe.

Southwest Asia — historical surveys, severe current uncertainty. Afghanistan Cannabis Survey 2010: ~40,000 growing households (25,000–60,000 interval) (Provenance A4). Resin-yield economics drove “largest producer” headlines specific to survey periods. Post-2022: opium contraction documented; no direct evidence establishes cannabis trajectory. Lebanon’s Bekaa: documented traditional resin region; figures dominated by eradication/seizures, not repeated crop surveys — on the map, not in a league table.

Europe — moderate for system/supply direction, low for total mass. EUDA: most EU herbal cannabis likely produced inside the EU, often indoors; Morocco principal external resin source. Western Balkans: herb seizures fell from >84 t (2017, largely Albania-linked) to ~19.5 t (2019) and 14 t (2021) — detected supply and operational change, not production totals.

Sub-Saharan Africa — broad presence, poor quantification. Cultivation/seizures across the continent; South Africa, Lesotho, Zimbabwe, Malawi, Nigeria combine traditional/unlicensed production with emerging medical/export licensing. INCB: 12 African countries submitted 2024 medical/scientific estimates — treaty channels, not total agriculture. No defensible ranking of African producers by total drug-type output.

South & Southeast Asia — historically important, weakly measured. India, Nepal, Thailand, Laos, Cambodia, Indonesia, Philippines: documented traditions; uneven administrative visibility; product definitions blur wild/feral, low-THC, licensed medical, and illicit flower. Thailand’s regulatory volatility makes year-specific comparisons fragile (policy → R03).

Oceania — bounded legal exceptions, incomplete prevalence. Australia’s medical sector measurable via licensing/trade; national illicit production is not. ACT: two plants/adult, four/household, artificial-light and hydroponic excluded — legal capacity, not uptake.

B. Systems and scale structure

Three environments, different visibility

Open-field dominates best-documented traditional resin regions and likely much unlicensed bulk herb (capital economics). Greenhouse/light-dep especially visible in regulated horticultural markets. Fully indoor strongly represented in North American licensed markets and European detected production — but no global dataset measures its share. “Most legal cannabis is indoor” is unsupported globally. Canada counterexample: 1.454M m² indoor ≈ 145 ha floor vs. 596 ha outdoor — but areas cannot be converted to output (crop cycles, utilization). Provenance B1 tabled.

Scale categories cannot travel cleanly

“Craft/smallholder/industrial” lack global thresholds; defensible descriptors are measurable (canopy, hectares, harvests, labor model, household dependence, capital intensity, market channel). License counts overstate active producers (dormant, stacked, provisional, multi-premises). California DCC explicitly documents harvest-data corrections, including thousands of erroneous 2021 records where each plant was entered as a separate harvest — track-and-trace is not frictionless truth.

Cultivation as livelihood

Rif and Afghanistan: household plots in marginal rural economies connected to brokers and resin processing. Regulated North America: home gardens → small farms → greenhouse campuses → warehouses. European illicit: domestic rooms → industrial indoor. Structures coexist; legalization does not swap one universal type for another.

Home cultivation: rules are easier to count than growers

Canada surveys: ~10% of consumers grew or had someone grow (2019 NCS); Statistics Canada synthesis: cultivation as a source for 14% of consumers in 2020, up from 8% (2018) and 9% (2019) — consumer shares, not household prevalence or mass (Provenance B2). Uruguay: IRCCA registry of growers and clubs (clubs 15–45 members, 99 flowering plants); registration counts understate unregistered cultivation. Malta and Germany allow bounded home-grow; robust uptake estimates not found. US states: definitions vary too much for a single comparable figure.

C. Traditions and landraces

What “landrace” can and cannot mean

A defensible landrace claim needs a documented farmer-maintained population, regional accession, or population-genetic evidence; a geographic name on a seed package proves marketing provenance only. Sawler et al. (2015): genotyped 81 marijuana + 43 hemp samples; only moderate correspondence between reported sativa/indica ancestry and genetic structure; strain names often not meaningful genetic entities. Ren et al. (2021): whole-genome analysis supports early East Asian domestication and basal/hemp/drug-type groups; sampling gaps limit reconstruction; does not authenticate products sold as “Thai/Afghan/Colombian.”

Regional traditions: evidence tiers

Moroccan Rif — farmer population and historical production: strong; current genetic purity: weak. Documented hybrid replacement with substantially higher resin yield/potency — the clearest documented case of market-linked displacement pressure on a named farmer population; establishes erosion/replacement in sampled communities, not extinction everywhere. Hindu Kush/Afghanistan — regional history strong; named commercial accessions variable; commercial labels not self-validating. Lebanese Bekaa — regional tradition moderate-strong; population-genetic monitoring not found. Jamaica, Thailand, Colombia, Mexico, Malawi, South India — cultural/historical evidence variable; contemporary geographic-name products are marketed products absent collection history or genotyping.

Is global landrace erosion documented?

Plausible, not yet measurable as a global rate. Rif: direct field evidence. Breeding reviews: extensive crossing, reduced documented germplasm. Genomics: names unreliable, repositories incomplete. Supports a conservation concern, not a quantified worldwide loss percentage.

D. Cultivation-facing market structure

Canada: administrative clarity with boundary limits

Health Canada monthly series (Dec 2024): packaged dried production 9,779,211 units; licensee packaged inventory 20,568,017; distributor/retailer inventory 16,731,290; medical sales 221,276; non-medical sales 12,119,501 units. “Unit” = package count, not grams (Provenance D1). Area series shows contraction from peaks, especially outdoors — consistent with consolidation; causation → beyond R02. Medical personal/designated registrations: 15,767 (Dec 2024) → 11,304 (Mar 2025), −28% — registration activity, not proof growing stopped.

California: more production, uncertain unlicensed denominator

2024 official market outlook: licensed production increased (track-and-trace + tax/market data); annual changes in illicit production “difficult to determine”; eradication suggests continued unlicensed cultivation without a census. Dashboard publishes correction history — trust the revised series; license count ≠ active farms; wet harvest ≠ retail dry flower.

Oregon and Colorado

Oregon: Metrc-derived harvest/transfer/price/sales for OLCC recreational only; persistent high inventories and falling wholesale prices = regulated oversupply within Oregon, not global overproduction. Colorado: state reporting with different definitions/windows. Cross-state rankings need harmonized dry-weight and active-capacity measures that do not exist.

Product allocation and trade

Administrative systems measure retail categories better than producer-level allocation (flower vs. extraction fungible; reprocessing). Legal international trade visible through INCB treaty reporting and national authorizations — INCB estimates are ceilings/requirements, not actual harvest or export mass. Canada, Portugal, Colombia, Israel, several African states participate; no single current globally complete product-harmonized actual-exports table found; customs codes can mix preparations, extracts, medicines, plant material; hemp can contaminate broad codes.

E. What is solid, what is estimate-theater

Evidence-quality ladder

Strongest: mandatory administrative data (licensed slice only). Strong but bounded: probability surveys and registries. Moderate: repeated crop surveys/field studies (good for place+period, risky projected forward). Directional: seizures, eradication, dismantled sites, stated origin. Weak: market forecasts, advocacy valuations, seed-catalog origin stories, press repetition, statistics with no reconstructable denominator.

Myth checks

“World’s most valuable cash crop” — Unsupported (advocacy calculations multiplying estimated plants/consumption by retail price vs. legal farm-gate values; no measured global quantity, consistent price level, or product boundary). The opposite claim not established either. “Morocco is the largest producer” — Unknown as stated (products, years, methods differ; no current all-product ranking). “Afghanistan is the largest producer” — historically period- and product-specific (resin, survey years); current status unknown. “Most legal cannabis is grown indoors” — Unsupported globally (no system-share dataset; Canada outdoor > indoor floor area; cycles prevent output conclusion). “Legalization eliminated illicit cultivation” — False as universal; persistence size varies and is often uncertain (CA outlook explicit; Canadian sourcing surveys measure consumption, not farms). Policy → R03. “Seizures show where most cannabis is produced” — Unsupported (joint outcome of supply, enforcement, routes, reporting; origin labels may record last departure point). “A license count equals growers” — Unsupported. “Acreage tells us production” — Unsupported without system and yield data. “Named landraces sold today preserve ancient regional genetics” — Unsupported without accession history or genotyping. “There are X million cannabis cultivation jobs” — Indeterminate (no global census with consistent labor definition; industry totals include retail/ancillary/induced jobs).

Data-gap map

Global area and dry-flower output → harmonized country reporting with drug-type/hemp separation, harvested area, cycles, dry mass, uncertainty. Licensed vs. unlicensed denominator → repeated supply-source surveys + production modeling with sensitivity ranges. System output shares → active canopy, cycles, saleable dry mass, facility classification. Africa/South/Southeast Asia → representative crop/livelihood surveys independent of seizures. Home-grow prevalence/output → comparable household surveys with identical definitions. Producer-level allocation → harvest-batch mass linked to product stream. Landrace status → georeferenced accession sampling repeated over time with community provenance. Legal trade → harmonized actual mass by material, destination, purpose. Active producer counts → facility-level active status, ownership linkage, canopy used, harvest activity.

F. Corpus linkage layer

No defensible global production denominator → R05-C031 (extends worldwide). Area ≠ output without cycles/yields → R05-C031, R05-C002 (confirms). Administrative data cover only regulated slices → R01-C004/C005 (confirms measurement-population lesson). Seizures/eradication are detection indicators → R03 (seeds). Morocco’s hybrid shift breaks historical resin-yield factors → R09–R12 (seeds). Strain names lack stable genetic identity → R09–R12 (confirms genotype/chemotype discipline). Home cultivation measurable via surveys/registries → R21 (seeds). Active canopy/cycles/dry-mass standards → R22–R24 (seeds). Regulated oversupply is jurisdiction-specific → R24 (seeds).

Candidate claims for ledger review (26)

1. Cannabis cultivation is documented in at least 155 countries — geographic spread, not production mass. [UNODC 2026]
2. No current, methodologically consistent global estimate of drug-type harvested area or dry-flower output. [UNODC 2025/2026; EUDA]
3. Legacy global ranges were heterogeneous assemblies; not current point estimates. [UNODC/EMCDDA]
4. Canada Dec 2024: 1,453,902 m² indoor; 596 ha outdoor (licensed). [Health Canada]
5. Canada’s area figures cannot establish output shares (cycles, utilization, yields). [Health Canada; R05-C031]
6. ~14% of Canadian consumers reported self/proxy-grown cannabis as a source in 2020 (8% 2018; 9% 2019). [Statistics Canada]
7. Morocco 2003: 134,000 ha; 3,080 t potential resin — historical, not current. [UNODC Morocco Survey]
8. Rif hybrid replacement made kif-based resin-yield conversions unreliable. [EUDA; Chouvy & Afsahi 2014; Chouvy & Macfarlane 2018]
9. Morocco remains major EU resin source, but modeled production vs. seizures conflict blocks a confident current estimate. [EUDA]
10. Afghanistan 2010: ~40,000 growing households (25,000–60,000). [UNODC 2010]
11. Post-2022 evidence insufficient for Afghanistan cannabis trajectory. [UNODC 2023–2026]
12. Most EU herbal cannabis likely produced within the EU; Morocco principal external resin source. [EUDA]
13. Western Balkan herb seizures fell sharply 2017→2021; seizure change ≠ production change. [EUDA]
14. No defensible global indoor/greenhouse/outdoor share estimate. [UNODC; EUDA; national]
15. California 2024 outlook: licensed production up; illicit changes difficult to determine. [CA DCC 2025]
16. California harvest data required substantial outlier corrections. [CA DCC dashboard]
17. Oregon’s series describes OLCC-regulated production only. [OLCC]
18. License counts ≠ active producers. [state administrative definitions]
19. INCB estimates are controlled-system requirements/ceilings, not actual totals. [INCB]
20. Commercial strain names frequently fail to represent stable genetic identities. [Sawler et al. 2015]
21. Genomics supports complex domestication; cannot authenticate products from geographic names. [Ren et al. 2021]
22. Landrace replacement evidence strongest in the Rif; no quantified global erosion rate. [Chouvy & Afsahi 2014; Clarke & Merlin 2016]
23. “Most valuable cash crop” unsupported — denominator/price basis incomparable with crop statistics. [claim-lineage audit]
24. “Morocco/Afghanistan largest producer” not stable — mixes herb/resin, years, methods. [UNODC; EUDA]
25. No globally comparable cultivation-employment total; repeated figures exceed the cultivation boundary. [source audit]
26. A defensible global picture requires harmonized boundaries, active harvested area, cycles, dry mass, explicit uncertainty. [synthesis]

Strongest sources (with reliability notes)

UNODC WDR 2026 data portal (https://data.unodc.org/wdr2026?page=0) · UNODC WDR 2025 statistical annex · Health Canada cannabis market data + licensed area (canada.ca) · Statistics Canada, Research to Insights: Cannabis in Canada 2023 · California DCC Market Outlook 2024 (cdn.cannabis.ca.gov 2025 PDF) · CA DCC Data Dashboard Glossary/Notes · OLCC Marijuana Market Data · EUDA EU Drug Market: Cannabis — Production; Trafficking & Supply · UNODC Afghanistan Cannabis Survey 2010/WDR2011 (unodc.org PDF) · Chouvy & Afsahi 2014, IJDP 25(3):416–423, DOI 10.1016/j.drugpo.2014.01.001 · Chouvy & Macfarlane 2018, IJDP 58:85–91, DOI 10.1016/j.drugpo.2018.04.013 · Sawler et al. 2015, PLOS ONE 10(8):e0133292, DOI 10.1371/journal.pone.0133292 · Ren et al. 2021, Science Advances 7:eabg2286, DOI 10.1126/sciadv.abg2286. Additional: Clarke & Merlin 2016, Crit. Rev. Plant Sci. 35:293–327, DOI 10.1080/07352689.2016.1267498 · Chouvy 2024, IJDP 129:104479 (landraces/terroir), DOI 10.1016/j.drugpo.2024.104479 · INCB Narcotic Drugs Estimates · IRCCA (vías de acceso; adult-use data to 31-12-2025) · EUDA resin-market changes · UNODC WDR 2018 Booklet 3 Cannabis · INCB Annual Report 2023 Africa; 2020 Oceania.

Limitations

Structured global evidence review, not a new crop census. Language/indexing bias toward English/French/Spanish institutional sources. Dashboards change after publication; figures date-stamped to source state at cutoff. No inference of cultivation from consumption prevalence; no package-count→mass, canopy→yield conversions; no averaging of incompatible estimates. Absence of a defensible figure ≠ absence of cultivation.

Search-stopping rationale

Concluded when four lanes reached diminishing returns: global institutional sources kept returning presence/seizures/heterogeneous figures rather than a harmonized denominator; regulated-market searches converged on the same administrative datasets and their published limits; traditional-region searches produced strong historical anchors without a current comparable ranking; landrace searches converged on a small set of genomic studies and field evidence, with most region-specific claims lacking accession-level verification. Remaining uncertainty is structural and recorded in the gap map.

ChatGPT deep research | 31 Aug 2026