R04 Verification B — Claim support (magnitudes), with quoted passages
Date: 2026-09-06 · Sources of record: OpenAlex abstracts (26 papers); full text — Dillis 2026 (PLOS Water), Bauer 2015 (PLOS ONE), Desaulniers Brousseau 2024a (PMC11253875), Wang 2019 (ACP), Westmoreland & Bugbee 2022 (Frontiers), ETCC project page, DCC rulemaking page, DPR ENF 24-19, 4 CCR §17223 (LII); secondary confirmation via web search for Wang 2017 per-unit-area multipliers and Mills 2012 headline figures. Blocked: Springer/BMC full text (Cloudflare), ScienceDirect (403), T&F (403), Wiley (paywall). PDFs from IERC (Medel) and Berkeley CRC (Carah) were not text-extractable.
B1. ChatGPT records (H1 Q1–Q13 + selected K)
| Record | Claimed | Source passage | Verdict |
|---|---|---|---|
| Q1 | Mills 2012: ~1% of national electricity, historical model | Secondary reporting of Mills 2012 (Semantic Scholar/Cary Institute summaries): “20 billion kilowatt-hours… 1% of all electricity used in the United States… six billion dollars annually.” Elsevier abstract not fetchable. | VERIFIED (secondary) |
| Q2 | 22.7 L/plant/day is an HGA 2010 draft-ordinance assumption | Bauer 2015 Introduction: “Marijuana is a high water-use plant, consuming up to 22.7 liters of water per day. In comparison, the widely cultivated wine grape… uses approximately 12.64 liters of water per day.” Ordinance provenance in Methods (R05-C014 already verified). | VERIFIED |
| Q3 | Carah 2015: ~430 vs 271 million L/km²/season cannabis vs grapes | Berkeley CRC PDF not text-extractable; Bauer confirms the per-day chain the comparison rests on. | PARTIAL |
| Q4 | Dillis 2019: 901 enrolled reporting farms; 58.2% report wells | Abstract: “annual reports, covering the year 2017, submitted by 901 cannabis cultivators… wells (58% of sites)… surface water diversions (22%)… spring diversions (16%)… wells more than twice as frequently reported among compliant sites (68%)” | VERIFIED |
| Q5 | Dillis 2026: cannabis top user in 71/4,367 catchment-months (1.6%) | Results: “Of all catchment-months sampled (n = 4367), there were only 71 (1.6%) in which cannabis was the top water user. In contrast, residential use (n = 1512; 34.6%) and non-cannabis agriculture (n = 1299; 29.8%)…” Also: “cannabis alone… at least a 1% impairment (n = 101, 2.3% of catchment-months)”. | VERIFIED verbatim |
| Q6 | Büser 2025: 231.23 vs 188.19 L/m²; 18.6% reduction | Abstract: “SDI resulted in a reduction of irrigation water usage by 18.6% compared to DI… weed dry biomass was reduced by 93.2%… inflorescence yield increased by 5% and CBD concentration by 9%.” Exact L/m² in full text (blocked). | VERIFIED (18.6%) / PARTIAL (L/m²) |
| Q7 | Westmoreland & Bugbee 2022: 3× P input → 12-fold leachate P; no yield/quality effect | Abstract: “Leachate P increased 12-fold in response to the 3-fold increase in P input… no significant effect of P concentration on flower yield or cannabinoid concentration.” | VERIFIED verbatim |
| Q8/Q9 | 46/58 (2012); 86/101 (2015) fishers exposed | S-158 (R03 verified). | VERIFIED (corpus) |
| Q10 | Brousseau 2024a: media 65–75% of GWP; 60.7 → 52.7 kg CO₂e/100 g THC; ReCiPe midpoint hierarchist v1.13 | Results: “Potting media contributed between 65 and 75% of the GWP impacts in both treatments.” “Using the fertilizer treatment H– to produce 100 g of THC has a GWP of 60.7 kg CO2-eq… on-site potting media reuse via steaming decreased the GWP… to 52.7 kg CO2-eq.” Methods: “OpenLCA v2.0… ReCiPe midpoint (hierarchist v.1.13).” Per-kg values: “H– … 61.8 kg CO2-eq per kg of dry flower… L+ … 110.7.” | VERIFIED verbatim |
| Q11 | Das 2026: ~6% cultivation-inclusive GWP reduction; ~70% drying-only | Abstract: “In the drying stage, CP-CED reduced environmental impacts to approximately 50%… MI-CED achieved… around 72%. In terms of global warming potential, CED contributed 11.31 kg CO2 eq. whereas CP-CED and MI-CED reduced this to 5.68 and 3.25.” Cultivation-inclusive 6% is in full text (T&F 403). | VERIFIED (72%) / PARTIAL (6%) |
| Q12 | Denton 2025: 377 mm ET | Abstract truncated at the results number; confirms eddy covariance, 2,000 plants/ha, “high water use and low water productivity.” | PARTIAL |
| Q13 | ETCC: 70% of water transpired to condensate | Project page: “70% of water use in CEH facilities is transpired to the room and ends up as HVAC condensate… estimated at 17.8 million gal/yr statewide in California.” | VERIFIED verbatim |
| K14 | Hofstadter 2021: WUI association, cannabis link unresolved | Abstract: “high rates of exposure (62%)… we found no clear link between illegal cannabis cultivation and AR exposure. However, Barred Owls sampled in proximity to the wildland–urban interface (WUI) were more likely to be exposed.” | VERIFIED |
| K18 | Urso 2022: measured exhaust inputs → small modeled regional effect | Abstract: “biogenic VOC emission rate from the sampled indoor facilities to be low (2.13 lbs to 11.12 lbs of VOC/ton of cannabis harvested)… exhaust profile lacked isoprene.” | VERIFIED |
| K23 | DCC-2025-01-R effective July 1, 2026 removes electricity reporting/offsets | DCC page: “Approved and filed with Secretary of State on April 28, 2026. Effective July 1, 2026… Deleting the requirement to submit electricity reporting with a renewal application and to purchase carbon offsets.” | VERIFIED verbatim |
B2. Gemini magnitudes
| Item | Claimed | Source passage / finding | Verdict |
|---|---|---|---|
| CC2, H19, M2 | Outdoor 22.8 kg CO₂e/kg; joint = 0.023 kg = 0.057 mi | Brousseau 2024a Results: 61.8 (H–) and 110.7 (L+) kg CO₂e/kg dry flower; 60.7 kg per 100 g THC; abstract: “50 times less carbon-emitting than indoor.” | FAIL — value not in source; downstream arithmetic wrong |
| CC19, H19, H20 | 50/50 rule forces landfilling; composting prohibited | §17223(b): “(1) On-premises composting of cannabis waste. (2) Collection and processing… organic waste collection route. (3) Self-haul… composting facility… in-vessel digestion… (4) Reintroduction… tilling directly into agricultural land and no-till farming.” | FAIL |
| CC15 | Zero synthetic pesticides permitted; 25(b)/sulfur/biologicals only | DPR ENF 24-19: legal if actives “exempt from residue tolerance requirements and from registration OR exempt from residue tolerance… and use would not be… in conflict with the registered label”; “Products exempt… under FIFRA 25(b)… will not be included in the list.” | OVERREACH |
| M7 | 0.1–0.3 ppb additional ozone; <1–2% of biogenic VOC | Wang 2019 ACP abstract/§3.3.1: “+362 t/yr… increases of up to 0.34 ppb in hourly ozone… morning and 0.67 ppb at night… every 1000 t/yr… 1 ppb… MDA8 increase of 0.3 ppb.” Inventory: base case 731,442 t/yr VOC, 47% BVOC; cannabis 66–657 t/yr. | UNDERSTATED; % unsourced |
| M10 | ~21% organic yield deficit | S-099 full-text (R08): mineral 23.7 vs organic 19.6 g/plant → 17.3%. | FAIL (wrong number, wrong system label) |
| CC7, H19 | 70–80% condensate recovery | ETCC: 70%. R05-C007 note: “Gemini’s 80–95% recapture figure unsourced; ledger takes the measured 70%.” | 70% VERIFIED; 80% unsupported (repeat) |
| M1 | 20 TWh; $3B; 1% | Secondary: 20 TWh, 1%, $6 billion. | 20 TWh/1% VERIFIED; $3B FAIL |
| Top-15 #7 | 428 cultivators (Dillis 2019) | Abstract: 901. | FAIL |
| CC6 | 58.2 / 21.6 / 16.2% wells / surface / springs | Abstract: 58 / 22 / 16% of sites. | VERIFIED (rounded) |
| CC10 | 1.5× more forest edge; mean clearing 0.16 ha; 68 watersheds | Abstract: “cannabis leads to far greater changes in key metrics on a per-unit-area basis.” Secondary (Anthropocene/press): “1.5 times more forest loss and 2.5 times greater fragmentation” per unit area. 0.16 ha not verified. Butsic 2016 abstract: “4428 grow sites in 60 watersheds.” | PARTIAL (multiplier direction ok; metric label and 0.16 ha unverified; 68 → 60) |
| CC12 | 79–85% fishers | S-158: 79% (2012, 46/58); 85% (2015, 86/101). Split by paper. | VERIFIED (must split) |
| CC13 | 70% NSO / 40% barred owls | S-158 (Gabriel 2018). Franklin 2018 = single carcass (“we report on an observation of brodifacoum AR exposure in… the northern spotted owl… found freshly dead”). | VERIFIED; Franklin is a case note, not prevalence |
| CC14 | 11% of monitored stations; six-year deployment | Medel 2022 abstract: detections “downstream of illegal public land cultivation sites… within four National Forests… no positive detections… within higher-order catchments”; no station %. | UNVERIFIED |
| CC5, H19 | 5.5 / 3.8 L/m²-day medians; 766 validated logs | Dillis 2020 abstract silent on medians; full text blocked. | UNVERIFIED — request page/table |
| CC16 | 77 / 165 / 388 kg N/ha; >200–388 kg N/ha | Brousseau 2024a text reports treatments H–, L+ and K/N ratios; no kg N/ha rates found. | UNVERIFIED |
| Line 255 | 2.8 L/plant-day; 380 L/kg (Brousseau 2024a) | Not found in PMC text. | UNVERIFIED |
| CC20, H20 | Packaging <1–3% of cradle-to-grave GWP (Brousseau 2024b) | Abstract: attributional production LCA, six categories; no packaging term; not cradle-to-grave. | UNVERIFIED; boundary mislabeled |
| CC18, H19 | LED: −30–40% fixture; −10–25% net facility | S-025: efficacy 1.7 (HPS) → 2.5 μmol/J (≈32% less electricity per photon). ETCC page silent on net facility %. | 30–40% supported by inference; 10–25% unsourced |
| H19 | Indoor gross water 6.1 (4.0–8.5) L/m²-day (ETCC) | Not on project page. | UNSOURCED |
| Line 657, H19 | 30–50% of injected CO₂ vented | No source given. | UNSOURCED |
| Line 176 | ~1.1% of national industrial-sector GHG | No source given; Mills 2025 abstract compares to cars/homes/crops/data centers/crypto, not an industrial-sector share. | UNSOURCED |
| M3 | 400–700 mm cannabis; 350–600 grapes; 900–1,200 alfalfa; 1,000–1,300 almonds | No source given. | UNSOURCED |
| CC11 | 2,500–3,200 acres licensed canopy; 24.3 M acres CA ag land; <0.013% | No source given; DCC publishes license tiers, not planted canopy; web search finds no statewide canopy total. | UNSOURCED (manufactured denominator) |
| Line 325, H19 | Bauer: 23% to 100% of streamflow | Results: “Upper Redwood Creek… 2–23%… Redwood Creek South… 34–165%… Salmon Creek… 36–173%… Outlet Creek… greater than 100%… but we cannot determine by how much.” | VERIFIED (as a paraphrase) |
| Exec / M1 | 43.7 Mt CO₂e; 595 PJ; half legal | Mills 2025 abstract: “∼44 Mt CO2e/year (half from legal producers)… 595 PJ/year… $11 billion/year… up to 76% by shifting more cultivation outdoors.” | VERIFIED |
B3. Inherited-claim calibration finding (for R05 audit)
R05-C013 reads “diversions could reach up to ~23% of seven-day low flow in studied watersheds.” Bauer 2015’s 23% is the upper bound of the least-impacted watershed; in two of four watersheds modeled demand exceeded the annual seven-day low flow (34–165%, 36–173%) and in the fourth exceeded a near-zero minimum by an indeterminable margin. The abstract itself says: “an estimated flow reduction of up to 23% of the annual seven-day low flow in the least impacted of the study watersheds. Estimates from the other study watersheds indicate that water demand for marijuana cultivation exceeds streamflow during the low-flow period.” Proposed correction logged in critique §5; not applied to R05 pending a second check (rule 10).