The Environmental Footprint of Cannabis: A Map, Not a Number

R04 Verified Synthesis — Wave 1, final curriculum slot

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

The question, and why it has no single answer

Ask “what is the environmental footprint of cannabis?” and the honest reply is a question back: which footprint, measured on what, for which cannabis? Every published figure in this literature is specific to a system boundary (operational electricity only, cradle-to-farm-gate, or further), a functional unit (a kilogram of dried flower, a gram of THC, a square metre of canopy for a year, a plant, a dollar), an impact category and the method that characterized it, and a production system grown by a particular population. Figures can be compared only where those four line up; the system and the population are what is being compared, not conditions to be matched away. The corpus’s deliverable is therefore the map, not a pooled number [R04-C001]. This cycle supplied its own cautionary tale: the one attempt to convert a per-kilogram outdoor figure into “a joint equals 300 feet of driving” used a number that does not appear in the paper it cited.

The map is smaller than the discourse suggests. The located life-cycle literature amounts to one modeled national indoor assessment, one Canadian industry model, one outdoor field-data LCA of a container system in Quebec, a 2026 drying study that borrows that Quebec inventory, a Washington master’s thesis, and adjacent hemp and prototype work. No LCA of a supplemental-light greenhouse, a light-deprivation hoop house, or in-ground full-sun drug-type cannabis was located within the documented search [R04-C002]. And the papers that do exist lean on each other: shared authors and inventories, one water-reporting lineage, overlapping wildlife samples, the same 2019 Colorado field campaign, and the same background emission factors, so agreement among them is correlated rather than corroborating [R04-C003].

Energy and carbon: inherited, traced, not re-litigated

R05 already established the indoor-versus-outdoor carbon spread and the rule that its figures cannot be averaged across boundaries. R04 adds provenance and context rather than new totals. The claim that cannabis consumes “one percent of US electricity” is a 2012 model built on 2011 assumptions, later echoed by an industry report that added voluntary facility reporting and market extrapolation rather than metering; no current national electricity share has been established, and the sector-scale energy and emissions figures from 2025 measure something else on a different year [R04-C004]. The joint-equals-driving equivalency has no universal value. On stated inputs the arithmetic can be shown for an indoor gram and for an outdoor-container gram, and the two differ by more than an order of magnitude, which is the point: the equivalency is a property of the boundary and the grid year, not of the joint [R04-C005].

The 2025 and 2026 publications widen the cases without changing the inherited range [R04-C006]. One of them isolates drying as its own endpoint: pretreatment cut drying-stage emissions by roughly half to nearly three-quarters, while the cultivation-inclusive reduction was far smaller, a distinction that must survive any retelling [R04-C007]. Cross-crop comparisons hold only on aligned units; dried flower against fresh produce by mass is a distortion, and no defensible per-dose or per-dollar comparison with other consumer goods was verified [R04-C008].

Water: from lore to a scaled answer

The most repeated water number in cannabis, 22.7 litres per plant per day, is a 2010 county-ordinance assumption that Bauer and colleagues adopted for modeling, and the “twice the water of wine grapes” comparison descends from that same assumption through an area-season extrapolation. It is a chain of assumptions, not a paired experiment [R04-C009]. Actual measurement exists only at experimental scale: in one tunnel trial, subsurface drip used 18.6% less seasonal irrigation than surface drip while yield and CBD were higher [R04-C010]; one eddy-covariance study measured field evapotranspiration of low-density floral hemp, which includes soil and interrow losses and cannot validate any per-plant figure [R04-C011]; indoor deficit irrigation maintained yield in one experiment, at leaf scale rather than facility scale [R04-C012].

Where the water comes from is documented administratively. Among 901 enrolled Northern California cultivators, wells were the most frequently reported source, then surface diversions and springs, with wells far more common among compliant sites; these are frequencies of sites, not shares of volume [R04-C013]. Stored water shifts extraction away from the season of demand but was generally insufficient for full forbearance, and since 2019 California has required forbearance from surface diversion through the dry season [R04-C014]. Modeled dry-season extraction is consistently greater for unpermitted than permitted farms, and scenario modeling shows how much permitting and storage could reduce it [R04-C015].

Then the scaled result. Across fourteen upper watersheds, cannabis was the top water user in 1.6% of catchment-months, against roughly a third for residential use and a third for other agriculture. Modeled potential streamflow pressure from residential and agricultural use appeared at annual scales; cannabis-associated pressure appeared only at monthly scale, chiefly in dry years. Cannabis water use is a spatially and temporally concentrated pressure, additive to larger users, not a uniformly dominant one [R04-C016]. That sits beside, not against, the older modeled finding that in four specific watersheds demand exceeded low-season streamflow in three, a range the corpus had understated until this cycle corrected it [R04-C017]. Indoors, about seventy percent of applied water leaves as recoverable condensate in the one metered study, and recoverable is not reused; no complete commercial water balance was located [R04-C018].

Land, habitat, wildlife: real harms, bounded populations

Systematic mapping of Humboldt County found thousands of grows, clustered, often remote from roads, often on steep slopes, and sometimes near threatened-fish habitat; these are siting proxies, not measured erosion or mortality [R04-C019]. Compared with timber harvest, cannabis converted less land in total but caused more forest loss and fragmentation per unit area [R04-C020]. What the literature cannot supply is a denominator: license tiers are maxima, no statewide planted-canopy total is published, and any “cannabis is X percent of farmland” figure is manufactured [R04-C021]. Post-legalization spatial associations between licensed presence and less unlicensed cultivation are observational, not proof that licensing protects habitat [R04-C022].

Within the literature located for R04, the strongest direct evidence of wildlife toxicological harm comes from trespass grows on public land: rodenticide exposure in most tested fisher and spotted-owl carcasses, diagnosed poisonings, reduced fisher survival near grow sites, and banned carbofuran, with the denominators being tested animals rather than wild populations and the trespass share of production unknown [R04-C023]. The counterweight is a barred-owl study in which most birds were exposed with no clear link to cannabis and a strong association with the wildland–urban interface, leaving dwellings and grows as competing sources [R04-C024]. Private-farm wildlife effects are located but unresolved: local activity responses from one shared dataset, and experimental light-and-noise work sitting unextracted in a dissertation [R04-C025].

Chemicals and air: what is measured, and where

Pesticide fate in the environment has been measured only at trespass sites, where carbofuran and diazinon were confirmed in headwater streams below eradicated complexes and absent from higher-order catchments; no fate monitoring of licensed farms was located, and product-residue testing is not fate monitoring [R04-C026]. The regulatory instrument is routinely misdescribed: no pesticide is registered for cannabis, California admits products whose actives are tolerance-exempt and label-compatible, and permitted products can still fail residue tests, so both “zero synthetic pesticides” and “legal means clean” misstate it [R04-C027].

The only measured nutrient-loss study found that tripling phosphorus input raised leachate phosphorus twelve-fold with no yield or cannabinoid penalty in that trial; no field leaching study of commercial cannabis exists [R04-C028]. The outdoor LCA delivers the cycle’s most counterintuitive result: the higher-fertilizer regime had the lower footprint per kilogram, because it needed about half as many plants and about half as much potting media, and changing the functional unit to THC narrowed but did not reverse the ordering [R04-C029]. Potting media were 65 to 75 percent of that system’s warming potential, and a modeled steam-sterilization reuse scenario was a substantial lever in that system; indoors, media are a minor term [R04-C030]. Terpene emissions are measured and their regional ozone effect is modeled as sub-ppb in the studied Colorado scenarios, with measured exhaust rates low and no isoprene; regional air-quality significance is not established as large, while odor nuisance is a separate, real question [R04-C031]. CO₂ enrichment has no verified net-impact study [R04-C032].

Waste and governance: mostly unmeasured

No representative mass-flow inventory of growing media, plastics, lamps, plant waste or farm-gate packaging was located, and no transparent packaging LCA; the claim that packaging is the industry’s largest impact is unsupported, and so is any claim that it is negligible [R04-C033]. California’s waste rule expressly permits on-premises composting and tilling; the fifty-percent rule governs how waste is rendered unrecognizable, not where it goes, and the “forced into landfills” story is false for California [R04-C034].

Mitigation levers rank only within an impact category on a shared baseline. For carbon, several non-equivalent models agree that shifting production away from energy-intensive indoor systems produces large reductions, with effect sizes that cannot be pooled; for water, one measured irrigation lever and modeled storage and timing levers; for nutrients, a measured phosphorus reduction; for habitat and waste, nothing comparable. Cross-category tradeoffs are real, and no overall ranking is defensible without an explicit, sourced weighting [R04-C035]. Regulations establish requirements, not outcomes: California deleted its cultivation electricity-reporting and offset requirements effective July 2026, and no causal evaluation of any instrument’s environmental effect was located [R04-C036]. Certifications establish criteria and inspection, not measured footprint improvement [R04-C037].

The public claims, calibrated

“One percent of US electricity” is a 2012 model, unsupported as a current fact [C004]. “A joint equals X miles” has no universal value [C005]. “Twice the water of grapes” is an assumption chain, and the scale question has a scaled answer [C009, C016]. “Sun-grown is sustainable by definition” is unsupported as a label; media, nitrogen, timing and siting each cut against it [C017, C019, C029, C030]. “Illegal grows are the problem” and “legal grows are clean” both fail as blanket statements, because the trespass evidence is population-specific and the licensed population is unmonitored [C023, C024, C026, C027]. “Cannabis is carbon-negative” is not established for drug-type flower [C032]. “Terpenes are a regional air problem”: modeled ozone increments from cannabis terpene emissions were sub-ppb in the studied Colorado scenarios [C031]. “Packaging is the biggest impact” is unsupported within the documented evidence [C033]. “LEDs solve indoor carbon” remains unsupported as a complete solution [C006]. “Organic has a lower footprint” is not established by label and depends on the unit [C029].

Then vs. now: what the archive taught

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

What would change these conclusions

A system-resolved LCA of greenhouse, light-deprivation and in-ground outdoor cannabis [C002]; a complete, representative commercial indoor water balance [C018]; a field nutrient-leaching study [C028]; environmental-fate monitoring on licensed farms [C026]; a measured mass-flow inventory of media, plastics, lamps and packaging [C033]; a planted-canopy denominator that is not a license maximum [C021]; and any causal evaluation of a regulatory or certification instrument [C036, C037]. Until those exist, the responsible statement is the map, with its blanks labeled.

Changelog

• v1.0 (2026-09-07) — Danny sign-off after ChatGPT’s editorial pass (conditional approval, no structural or grading changes). Four calibrations applied: C010 wording bounded to “in one tunnel trial… used 18.6% less… than surface drip”; the archive-bridge sentence no longer assigns the metered 70% condensate coefficient to Danny’s historical facilities; C029 restored “about half”; the terpene public-claim line now says “modeled ozone increments… were sub-ppb in the studied Colorado scenarios.”
• v0.1 (2026-09-07) — initial draft from ledger v1.0 (ChatGPT-calibrated).