# Hamilton–Niagara Industrial Atlas A map of the **physical industrial economy** of Hamilton, Burlington and the Niagara Peninsula — where things are made, repaired, fabricated, stored and moved, and how rail, road, port and land fit together into one regional productive system. Not a business directory. The question it answers is *"what does the physical industrial economy of Hamilton–Niagara actually look like?"* — not "where are the businesses?" An application in its own right, living in this folder inside the Paton Hall site repo. It has no dependency on the site's build (`build.py` only reads `../src/`), and GitHub Pages will serve it at `/site/atlas/` once there is an `index.html` to serve. --- ## Status: reconnaissance and normalization complete The data-source reconnaissance pass is finished and the normalization pipeline is built. The unit index (445,075 units across three tiers) exists, carrying provenance on every feature. Enrichment, change detection, scoring, emit and the dossier remain to be built; nothing is rendered in the app yet. **Reference documents:** | Document | What it holds | |---|---| | [GLOSSARY.md](GLOSSARY.md) | places and terms — Niagara Region vs Peninsula vs study area | | [logs/](logs/) | raw findings log, appended as work happens | | [PUBLICATION-MODEL.md](PUBLICATION-MODEL.md) | who the Atlas is for, and how the work reaches them | | [ENVIRONMENT.md](ENVIRONMENT.md) | packages, dependencies, environments and each package's role | | [TECHNOLOGY-DECISIONS.md](TECHNOLOGY-DECISIONS.md) | why each tool was chosen over its alternatives; an evolving decision record | | [CANDIDATES.md](CANDIDATES.md) · [CANDIDATES-HAMILTON.md](CANDIDATES-HAMILTON.md) | the numbered dataset menus (#1–126 Niagara, #201–290 Hamilton) | | [SCHEMAS-HAMILTON.md](SCHEMAS-HAMILTON.md) | observed field-level schemas of fetched Hamilton layers | | [INGESTION-LEDGER.md](INGESTION-LEDGER.md) | what is held locally, verified when, pulled when | | [RECON-2026-08-22.md](RECON-2026-08-22.md) | municipal/provincial/federal reconnaissance and the government roster | | [DATA-SOURCES.md](DATA-SOURCES.md) · [BACKLOG.md](BACKLOG.md) | the 2026-08-18 baseline, and everything found but not loaded | **Done:** - [`DATA-SOURCES.md`](DATA-SOURCES.md) — every source verified against its live endpoint on 2026-08-18, with observed record counts, field names and licences. Gaps and dead ends recorded explicitly. - [`sources/sources.json`](sources/sources.json) — machine-readable registry, ready for the UI to expose so the atlas is auditable. - **Ingestion** — Hamilton and Niagara address points, parcels, footprints, and NEI business records cached. - **Normalization** — schemas unified across municipal/regional/national boundaries, NEI stabilized on `nei_id` rather than address, departures corrected (34 true departures, not 68), address anchor index (354,601 keys), parcel/footprint tiers (79,369 + 282,567 units), coverage per municipality. - **Unit index** — 445,075 units (parcel, footprint, address tiers) in UTM 17N, each with centroid, area (where applicable), and provenance block. Scoped to the twelve Niagara Region lower-tier municipalities plus Hamilton by an explicit allow-list; 11 units in stray regional/county labels (not a lower-tier municipality) are excluded and the exclusion count is reported, not dropped silently. **Not started:** enrichment and constraints, change detection, scoring, emit, the dossier, the Leaflet application, basemaps. ### What the reconnaissance established Three sources carry the first milestone, all verified live: | Source | What it gives | Verified | |---|---|---| | **OSM / Overpass** | Region-wide industrial land, buildings, rail, ports — one schema across every municipal and national boundary | 311 industrial-landuse ways in the Hamilton bbox | | **City of Hamilton Open Data** | Employment lands, zoning, building footprints, railways, truck routes | 463 datasets; 214,293 buildings; 11,931 zoning polygons | | **Niagara Consolidated NEI** | Business points with NAICS, sector, employee size range, indoor floor area | 98,065 records under OGL 2.0 | **The finding that shapes the build:** Niagara publishes a rich business inventory and Hamilton does not. The first version will therefore be asymmetric — business-level detail in Niagara, land- and building-level detail in Hamilton. That asymmetry has to be *shown*, not smoothed over, or a reader will conclude Hamilton has less industry than Niagara, which is false and is purely an artefact of who publishes what. --- ## Next step Per §20 of the brief, Layer 1–3 of the MVP, in this order: 1. **Ingestion** — `scripts/osm/` and `scripts/hamilton/` writing raw pulls to a cache, respecting rate limits. 2. **Normalization** — into the atlas schema, every feature carrying its full provenance block. 3. **A first map** — Leaflet over pre-generated GeoJSON, showing OSM industrial places, the rail network and Hamilton employment lands, with per-layer freshness and source attribution visible from the first commit rather than retrofitted. Directories get created when a real source justifies them, not up front. --- ## Rules this project runs on Carried from the brief, and worth restating because they are the difference between an atlas and a plausible-looking picture: - **Never fabricate data.** A gap is shown as a gap. - **Never infer that a business is currently operating** because an old dataset lists it. The NEI is a multi-year inventory; a 2016 row is a 2016 observation. - **Provenance is a feature, not metadata.** Source name, URL, retrieval date, licence and original ID travel with every feature. - **Distinguish observation from inference, and current from historical.** A railway from a government GIS file and a vessel seen four minutes ago are different kinds of claim and must never be rendered as the same kind. - **Every layer declares a freshness state** — `LIVE`, `RECENT`, `UPDATED MONTHLY`, `UPDATED ANNUALLY`, `HISTORICAL`, `STATIC`, `INFERRED`. - **Derived scores are analytical indicators, not economic truth.** Label them that way in the UI. - **Respect licences, rate limits and `robots.txt`.** Cache responses. The Seaway's transit page is public to read and `Disallow: /` to crawl; it is registered in `sources.json` with status `excluded` so nobody re-litigates it later. - **Do not expose sensitive infrastructure detail.** Use only what is intentionally published. ### Licence note The three primary sources are on three different licences — ODbL (OSM, **share-alike**), the City of Hamilton Open Data Licence, and OGL 2.0 (Niagara). Any derived layer that mixes them, such as the industrial density heatmap, raises a compatibility question that should be answered deliberately before publication rather than discovered after it. --- ## Reference implementation Architecture follows [`babbworks/atlas`](https://github.com/babbworks/atlas) (branch `master`) — a static Leaflet + OSM/Overpass application for UK industrial spaces. Its shape confirms the pipeline: a static frontend over **pre-generated JSON** built by Python ingestion scripts, with one module per source and a separate scoring module for derived indicators. It splits large datasets by key into many small files to keep fetches cheap — a pattern worth copying for Hamilton's 214k building footprints. Preserved from it: lightweight, modular, client-side, no mandatory database, statically deployable, independently replaceable data layers, GeoJSON-friendly.