Data credits

Data provenance and reuse status

Horticultural factual intake (2026-10-03)

The expanded top-500 pass collected factual fields from Lady Bird Johnson Wildflower Center and Missouri Botanical Garden Plant Finder. The combined Wildflower NC/TN distribution indexes contained 4,418 names, with 396 exact target matches; Missouri's native index contained 658 names, with 111 exact target matches. These are adapter-specific discovery results, not a claim that unmatched plants are absent from the entire websites.

The expanded pulls retained 2,318 growing-feature records and 506 bloom-month records across 507 source profiles. Relative to the pre-expansion top-500 snapshot, effective coverage gained 126 height, 44 spread, 58 sunlight, 44 moisture, 183 flower-color, 72 hardiness, 101 leaf-habit and 298 life-cycle fields. Existing source guidance, Quercus scores and Fit inputs were preserved; 27 conflicting fields remain withheld alternatives. Structured labels and straightforward qualified height statements are parsed by scripts with zero model extraction calls. Narrative descriptions and photographs are not republished by this workflow. Source URLs, snapshot hashes, retrieval dates and explicit scope remain attached; imports reparse the archived source to verify incoming values.

Bloom months remain in data/plants/regional-phenology.json, with unspecified geography where the page does not establish a reporting location. A plant's NC distribution does not make its reported bloom window NC-specific. Bloom modeling remains a separate workstream. Collection caches and full intake records are not published. At the end of this first expansion, 238 of the top 500 lacked height, sunlight or moisture. These additions are staged locally and have not yet been published to D1.

The subsequent targeted gap search added 240 selected source fields across 71 catalog plants. It used NC Extension's complete sitemap, Missouri genus searches outside the native-only index, Minnesota Wildflowers and existing unambiguous botanical aliases. Minnesota imports explicit height and life cycle only; its habitat descriptions are not growing tolerances. Wildflower height parsing now recognizes “to about” as a qualified upper bound. After this pass, the top 500 have 76 height, 187 sunlight and 200 moisture gaps; 209 plants lack at least one. Automated access failures and ambiguous identities are recorded separately. This pass is also staged, not live.

Full-catalog growing-data expansion (2026-10-03)

The full-catalog backfill processed the entire frozen missing-field queue, using NC Extension's sitemap and Minnesota Wildflowers' name index. It retained 1,413 verified NC Extension profiles and 885 Minnesota profiles, adding 8,558 selected growing fields across 1,722 catalog plants without replacing existing displayed guidance. Source-backed synonym links can share factual characteristics across preserved historical IDs; native status and scores are not transferred. Photographs are not imported by this workflow.

Full-catalog height, sunlight and soil/moisture-guidance gaps are now 2,235, 2,729 and 2,815 respectively. 2,995 plants lack at least one of these fields. These source facts are distinct from numerical Fit inputs. Downloaded profiles and intake records are not published. The additions are locally staged, not published to D1.

Tennessee staging (2026-09-26)

data/staging/tennessee/ is a separate, unpublished evidence package. Sources are the Tennessee–Kentucky Plant Atlas's public species/synonym/county search responses, UT Smart Yards page snapshots and TN Invasive Plant Council lists. Every downloaded source has request metadata and a hash in data/raw/tennessee-2026-09-26/manifest.json; the build also hashes its local inputs. Public access does not establish bulk redistribution rights. No third-party images were downloaded for this package.

County occurrence, regional native flags, explicit TN native assertions and TN invasive listings remain distinct. NC scores and geography are not transferred. The separately collected rare-plant PDF is the 2025 edition and is retained for review, not used to override atlas ranks. Production source loaders do not read this package.

Current North Carolina inputs

Inputs originate in the user-supplied quercus-v0.15 archive. The metadata below describes the supplied snapshot's attribution; it is not a newly granted license. Public method and species pages include source links. Do not apply a blanket code license to third-party datasets.

Full citation URLs are listed on the method page. Build inputs remain versioned with the source so results can be traced to a release commit. Personal yard records and raw examples are excluded from the repository and generated website.

The USDA shade-class interpretation is based on an inverse association in 193 exact-name woody-species matches: High=2.009, Medium=2.723, Low=3.469 on the independent 1–5 shade-tolerance scale. It does not identify the upstream cause or establish universal correctness for herbaceous plants.

Habitat categories and the habitat bridge (2026-10-19)

data/habitat/types.json adopted nine owner-approved types from the habitat bridge (Codex research, habitat-bridge-20261018): hardwood flatwoods, disturbed wetlands, sounds and tidal flats, tree plantations, Coastal Plain hammocks, shale barrens, Florida scrub, mangrove swamps and coastal tidal marshes whose salinity is not stated, plus Gulf state extensions of the coastal types. Descriptions are plain glosses awaiting owner review. data/habitat/class-types.json (tools/import_habitat_bridge.py) files USGS GAP land-cover classes (public domain), NatureServe ecological systems (CC BY 4.0) and 2011 NVC groups under these types: accepted bridge rows only, per registered state where both the class and the type occur. Provisional rows stay in the bridge for review. Association only; never a feeding link or a score input.

Bird ecology (2026-10-18)

Bird pages (data/wildlife/birds/, tools/build_bird_ecology.py) use:

• USGS Gap Analysis Project species habitat models (2001), https://doi.org/10.5066/F7V122T2, and species range maps (2001), https://doi.org/10.5066/F7Q81B3R. Public domain. Each model's own DOI is kept. Raw modeler notes are not published.
• NatureServe Explorer ecological systems and plant-association composition, CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), cited on every page that uses them. The bird release stores the plant names as NatureServe supplies them, with community and taxon UIDs as evidence; they are matched exactly and uniquely to the active plant publication when read (own scientific names before synonyms, joined taxa followed to their card), so a plant publication updates them without a bird rebuild. They show shared habitat, never food.
• US Census TIGER/Line 2025 counties (public domain), overlaid on GAP-modified HUC12 watersheds.
• Bird photographs from iNaturalist with photo-specific CC0, CC BY or CC BY-SA licences, artist and source kept on every image. No Cornell (All About Birds, Birds of the World, eBird, Macaulay, Merlin) content.
• Album additions (2026-10-19, data/wildlife-photos/birds/media.json): gap-fill iNaturalist photographs after a second full-size review (Male and Female as the source annotates them; Adult where no sex is claimed, never relabeled), then public-domain or CC0 prints as Illustration: Audubon (National Gallery of Art open access; Smithsonian Libraries, Internet Archive and University of Pittsburgh scans), Catesby (Smithsonian Libraries via the Biodiversity Heritage Library; National Gallery of Art), Gould and Fuertes. Nine Catesby plates bound sideways in the scanned copies are stored as lossless 90-degree rotations (quercus/jpeg_rotate.py), recorded with the original and derivative hashes. Print scans are published to the photo store from the research archive, bound by hash; they are not committed to the repository.

Caterpillar derived-data notice

Contains information from Shropshire & Tallamy (2025), ZooKeys 1261, made available under the Open Database License 1.0. Quercus filters host records to its NC reference cohort, consolidates genus names and counts recorded caterpillar species. The resulting caterpillar host-count data are made available under that same license in each public plant JSON record (the caterpillars factor). This notice does not apply a blanket license to other datasets or establish permission for sources whose terms remain unresolved.

Plant photo pilot (sandbox, 2026-09-26)

Ten Wikimedia Commons images, selected using Wikipedia species-page image associations and verified against Commons file metadata. Licenses are CC BY 3.0, CC BY 4.0, CC BY-SA 3.0 or CC BY-SA 4.0 as recorded per image. File descriptions, artist credits, license URLs, download URLs and SHA-256 hashes are preserved in data/plant-photos/manifest.json; original API snapshots in data/photo-pilot/. Photos are Commons-generated thumbnails, without local cropping or retouching; applicable image license and original-source links accompany each figure.

Reuse reference: https://commons.wikimedia.org/wiki/Commons:Reusing_content_outside_Wikimedia

Shared identity links determine which state pages reuse the photo; unresolved taxonomic matches do not receive photos by fuzzy matching. Whole-tree photos illustrate appearance and are not sufficient by themselves for identification.

Tennessee caterpillar scoring extract

data/states/tn/lep_hosts_tn.json derives Tennessee established-range host associations from Shropshire and Tallamy 2025, https://zenodo.org/records/17747343 (ODbL 1.0). The saved workbook hash and source IDs accompany the extract. Unique native Lepidoptera IDs are counted per exact host genus in the existing plant-only vocabulary, excluding the separate stray/fugitive range column. Applying the same extractor to NC reproduces all existing genus host counts. TN range overlap and a host association are not proof of a local feeding observation.

The derived tn-reference-1 assessments retain method 0.6.4's NC distributions as a fixed benchmark and use broader regional bee/bird sources with explicit scope labels. They are not ranks among all Tennessee species.

Tennessee reconciled evidence (tn-reference-2)

UT Tennessee Smart Yards collection context (https://tynnativeplants.wordpress.com/) and exact entries in the saved TDEC 2025 Rare Plant List provide state-specific claims. The derived layer retains original source text and scope. Atlas regional native status plus exact NC/USDA native evidence and TN occurrence is labeled as an inference, not confirmed indigenous county range.

The Avian Diet Database full source was retrieved for genera excluded by the older NC-genus filter (CC0-1.0, https://github.com/hurlbertlab/dietdatabase). SHA-256 a23130d880f0780faac9884acad300573fca1284dfd4aa7409458a10e0cfef61. Existing curated NC counts take precedence for previously covered genera. New records retain the same North American geographic scope, plant prey and exact-genus rules. TN atlas growthHabit fills missing USDA habit via a documented category mapping. Unapproved species synonym candidates do not transfer facts.

Current shared invasive scoring policy

TN-IPC listing determines Tennessee invasive status. Confirmed listed species reuse matched NC harm profiles and penalties; the transferred severity is explicitly a Quercus policy, not a TN-IPC rating. Unmatched profiles receive a provisional -30 Harmful baseline. Established/Emerging categories are not misrepresented as severity ranks. This supersedes earlier categorical-only Tennessee scoring notes. The method page describes the current shared policy.

Field-guide display enrichment

FloraQuest regional botanical intake — 2026-10-04

The user-purchased FloraQuest: Carolinas, Georgia Android package supplied a readable bundled SQLite database. Exact unique existing names/aliases matched 3,625 catalog identities. The source-specific adapter records habitat-light and habitat-moisture regimes for 3,571 plants, habitat and range accounts, original phenology, a 1–9 heliophily index, regional wetland indicators, botanical comments, and original identification attributes under floraquest_carolinas_georgia in data/plants/growing-traits.json. Database, record and field locators preserve provenance; category meanings are checked against the app's bundled definitions. Source habitat regimes do not become numeric Fit inputs or localized cultivation advice.

FloraQuest phenology follows Weakley's "flowering; fruiting" form ("Mar-May; Sep-Oct."). Where the text has exactly two parts and the second is a plain month span, the fruiting months fill fruit_season_begin/fruit_season_end (993 plants, quercus.growing_traits.fruit_from_phenology) and outrank USDA PLANTS' national fruit/seed period, which often starts at flowering. Next comes NC Extension's "Display/Harvest Time" read from Plant Toolbox pages already archived in this repository (tools/import_ncsu_archives.py, group ncsu_fruit_archive, 136 plants); USDA is the last fallback. The same tool adds full extension profiles for archived Extension pages that now carry a new name and list the catalog name as "Previously known as" (Fothergilla major, Hibiscus moscheutos, Coreopsis verticillata); the exact-name Extension crawls missed them.

29 explicit whole-plant height ranges from terminal species identification-key leads supplied 17 missing display heights. The subsequent display fallback fills 2,133 missing light fields and 2,212 missing soil/moisture fields with categories labeled “documented habitat,” preserving existing selected guidance and original source evidence. Species-linked key paths are separately queued for text review, with ancestor-group traits distinguished from terminal species statements. App photos were not imported. The purchased package, full database and raw review text are not published. This pass is staged, not published.

data/plants/growing-traits.json extracts factual growing observations from the existing USDA snapshot (983 species-name keys) and 164 exact species matches from a cached 326-plant priority NC Extension collection, including the earlier 24 pilot matches. Only structured field values are imported; narrative descriptions and images are excluded. Source URLs, retrieval dates and snapshot hashes are retained. Height ranges remain ranges, seasonal bloom values remain seasons, and drainage/pH categories are not converted into numerical calculator tolerances. Fifteen redirected taxa are held for identity reconciliation. The observations are displayed on matching plant pages and exposed in public plant JSON; Quercus scores and calculator inputs are unchanged.

The 2026-09-30 iNaturalist pilot adds 30 visually reviewed illustrations using official per-photo API license and attribution metadata. Only CC0 and CC BY photos are included. iNaturalist's license implementation, checked 2026-09-30, maps CC BY to 4.0 and CC0 to 1.0. Credits, source links, license links and unchanged medium-image bytes accompany each image. data/photo-expansion/inaturalist-pilot/release-evidence.json retains the per-photo API evidence, exact hashes and actual-image review notes. Research-grade identifications plus visual consistency review support illustrations; they are not expert taxonomic certification. Seventeen sampled identities remain held for diagnostic evidence and three name matches are unresolved.

data/plants/field-guide-details.json contains concise sourced facts for 20 identities from NC State Extension Plant Toolbox, NC State's herbarium, and the U.S. Fish & Wildlife Service. Each entry retains its source URL and review date. Month timelines are used only when explicitly supported; seasonal-only wording remains text. These facts do not modify calculator inputs or scores.

data/wildlife/wildlife-display.json indexes individual larval host names from the same Shropshire & Tallamy ODbL source and range filters as the existing genus counts, plus named bird associations from the Avian Diet Database (CC0). It is display enrichment, not a new scoring extract.

The photo register started with 53 approved images for 18 identities and is expanding across the shared catalog. Current approved coverage is recorded in data/photo-expansion/coverage.json. Each image carries its own source, photographer, reuse terms, modification notice and SHA-256. Initial independent decisions are in data/photo-review/review-decisions.json; subsequent decisions are in data/photo-expansion/reviews/. Source collection alone does not authorize publication: photographs require independent visual identity and rights review, and the promotion tool verifies the reviewed bytes. Rejected and unresolved candidates remain unpublished. Public images are resized source thumbnails, with no additional edits by Quercus. The public photo register is /photos/credits.json.

The 2026-10-01 iNaturalist catalog pass uses cached scripted collection, exact active taxon matches, and research-grade observations. Only individually licensed CC0, CC BY 4.0, and CC BY-SA 4.0 photographs are eligible. A single candidate per uncovered plant is the default, with compact independent contact-sheet review. data/photo-expansion/inaturalist-catalog/release-evidence.json retains accepted per-photo API rights, source taxon, reviewed hash, and visual decision. Research grade does not establish expert verification.

The October 2 bounded follow-up collects at most two new candidates for each of 50 previously uncovered plants, reusing cached observation responses and excluding their previous candidates. Independent Luna review assigns catalogue view categories. Promotion permits up to two distinct views per plant, preserves existing photographs, and checks the source rights and reviewed bytes. Holds and rejected photographs remain unpublished.

Supplemental growing facts — September 30, 2026

The October 1 Forest Service pass matched the full catalog against the official FEIS review index and Silvics chapter directory. It cached 474 matching reviews covering 411 identities, with raw hashes and section-bound excerpts. data/plants/forest-service-notes.json retains the scoped narrative evidence; only reviewed prose is published as growing notes. Explicitly reviewed botanical height descriptions fill missing descriptive fields through data/plants/growing-traits.json. Seedling responses, habitat associations, regional qualifications and other dimensions remain distinct. The original three-note pilot is retained as provenance.

EOL TraitBank's bulk archive supplied exact-name USDA PLANTS traits with original source URLs, trait IDs, maturity definitions and original units. Only confirmed foot-unit mature heights and bloom descriptions fill existing display fields; shade/drought tolerance and moisture use remain distinct source categories. SERNEC regional herbarium archives and BIEN supply source-scoped specimen habitats and height measurements in data/plants/plant-trait-evidence.json. Each retains attribution and evidence scope. Specimen notes and study maxima are not universal tolerances, native-range claims, or ecological-score inputs. No images were imported from these pulls.

The full-catalog import adds supplemental observations for 773 shared NC/TN plants.

Prairie Moon Nursery's 2026 Cultural Guide, Virginia DCR's Piedmont and Coastal Plain charts, Virginia Extension's plant table, and cached Tennessee Smart Yards pages were checked against the remaining priority plants. Imported records retain source URLs, PDF hashes and page coordinates, or cached page identity. Only factual categories and dimensions are extracted; photos and source prose are not reproduced. Prairie Moon heights describe plants in flower. Virginia DCR sunlight categories describe minimum requirements. Source categories remain descriptive growing observations, separate from ecological scores and calculator tolerances.

Yard planner (2026-10-01)

• Mature height, spread and flower colour come from the selected values already in data/plant-facts.json (USDA PLANTS, NC Extension Plant Toolbox and other recorded sources). The planner adds no new plant facts; sizes it cannot find are estimated in the browser from growth form and labelled as estimates.
• Optional soil lookup: USDA NRCS Soil Data Access (https://sdmdataaccess.sc.egov.usda.gov/), public domain. Queried from the visitor's browser only; Quercus does not proxy, receive or store the location.
• three.js r128, MIT License (vendored, license at /vendor/three-LICENSE.txt).

Garden suitability

data/plants/garden-suitability.json holds curated, source-linked planting cautions. Pest cautions quote FloraQuest's botanical comments on a pest or disease (hemlock woolly adelgid on both hemlocks, chestnut blight on American chestnut, balsam woolly adelgid on Fraser fir, butternut canker, Torreya canker, dogwood anthracnose, Dutch elm disease on American and slippery elm). They are marked informational: shown wherever the plant appears, never a reason to drop it from suggestions. Every other caution removes the plant from suggestions; quercus.garden_suitability.suggestion_excluded is the one rule, projected to browsers as suggestions_excluded. Contact and sap hazards, selected highly poisonous plants, and the requested very-weedy species are hidden from planting suggestions by default. An explicit checkbox includes them with visible warnings. Native status, wildlife scores, and growing-condition inputs are unchanged; existing yard plants remain visible with cautions. Species rules cover named infraspecific taxa and exact scientific aliases, not whole-genus weed judgments. Unlisted plants are not_assessed, not certified safe. Cautions distinguish contact, sap plus sunlight, and ingestion. The current list is an initial reviewed set, not a complete toxicology assessment of the catalog.

County outline display

quercus/assets/county-outlines.json contains the county (parish, independent city) outlines of every live state, written by tools/build_county_outlines.py from U.S. Atlas 3.0.1 counties-albers-10m.json (https://cdn.jsdelivr.net/npm/[email protected]/counties-albers-10m.json, sha256 a674dfa31b625e92635f32684a61ae135b05a91507f17f2d5164833237fecf46, retrieved 2026-10-06 and again 2026-10-19). U.S. Atlas redistributes the U.S. Census Bureau's 2017 cartographic boundaries, projected in Albers USA. TopoJSON delta arcs are decoded and rounded to two decimal places as SVG paths; the generator reproduces the earlier NC, TN, GA, SC and KY entries byte for byte before it writes. No county occurrence evidence comes from this file. The ISC notice is retained in county-outlines.LICENSE.txt and shipped with the map.

Southeast wave 1: Alabama, Florida and Virginia (2026-10-19)

Owner decisions of 2026-10-19 govern these sources. Staged evidence and receipts (raw API responses, PDFs, hashes) stay in the research folder (southeast-staging-20261019, outside the repository); tools/build_state_packages.py records the sha256 of every staged file it reads in each package.

The earlier staged GBIF API facet counts for Virginia (VA/gbif-open-specimen-counties.json) are superseded by the download and not used. CC BY-NC specimen data stays private evidence.