# subwaybuilder-demand-archive

Durable git-LFS archive of the **exact pinned raw source bytes** for the
[`subwaydemand`](https://github.com/) demand-data pipeline. This repo exists
because several vendors do not keep their files permanently — Geofabrik purges
dated `.osm.pbf` extracts, RLP LoD2 `/current/` rolls ~yearly, and BA directed
commuter data is interactive-only — so the pipeline's reproducibility (design
spec §3b) depends on these archived bytes rather than re-fetching from source.

**This is not the code.** The pipeline code lives in the `subwaydemand` repo.
This repo holds only raw data, tracked with git-LFS.

## Layout

`raw/` mirrors the pipeline's local `raw/` directory exactly, so restoring is a
plain copy (see below). `<region>.lock.json` snapshots are the pins these bytes
satisfy — every file's `sha256` in the lock matches the corresponding file here.

```
raw/
  bkg_lod2/   RLP LoD2 CityGML (geb3dlo) tiles for Mainz AGS 07315 + .meta4 + .txt sidecars
  zensus/     Zensus 2022 100m population grid (nationwide zip)
  osm/        Geofabrik dated .osm.pbf extracts: Rheinland-Pfalz + Hessen (2026-07-07)
  ba/         BA "gemband" svB xlsx (Gemeinde totals + commuter marginals)
  ba/interactive/  Browser-extracted BA snapshots (no stable vendor URL):
                   Regionalreport sector crosstabs (bst-reg, Mainz) + directed Pendler OD CSVs
rhein-main.lock.json   snapshot of the region lock (region RMN) these bytes satisfy
```

## Restore into the pipeline (precursor step)

From the `subwaydemand` repo root, with this archive cloned alongside:

```bash
git -C ../subwaybuilder-demand-archive lfs pull
cp -R ../subwaybuilder-demand-archive/raw/. raw/
uv run subwaydemand verify        # S0: sha256-gates raw/ against regions/<region>.lock.json
```

`subwaydemand verify` refuses to proceed on any missing/mismatched file, so a
clean verify confirms the restored bytes are bit-identical to the pins.

## Provenance

Verified acquisition notes (exact vendor URLs, formats, hashes) live in the
`subwaydemand` repo under `docs/research/`. Licenses: Zensus © Statistisches
Bundesamt (dl-de/by-2-0); RLP LoD2 © GeoBasis-DE/LVermGeoRP (dl-de/by-2-0);
BA © Statistik der Bundesagentur für Arbeit; OSM © OpenStreetMap contributors
(ODbL). Redistribution here is permitted under those terms with attribution.

Region **RMN** (Rhein-Main; v1 = Mainz), pinned 2026-07-08.