Asheville Land Map
Live at land.avlgo.com, part of AVL GO.
This is an Asheville, NC deployment of Civic Mapper by the
Center for Land Economics
(upstream repo, MIT licence, kept in
LICENSE). It adds:
- Asheville data:
data/jurisidictions/run_asheville.py builds the city's parcels from Buncombe
County's public parcel layer (2026 tax-year values, which still use the 2021 reappraisal).
Values are served as PMTiles from Vercel Blob.
- A parcel review list (
/review): parcels whose records look wrong, and the biggest
underused land. It's built by data/scripts/build_review_feed.py.
- AVL GO styling (
docs/avlgo-style-guide.md), with the site limited to Asheville.
General fixes found along the way are offered back upstream. Everything below is the upstream
Civic Mapper README.
Civic Mapper
Interactive 3D maps of parcel-level property data — land value, improvement
value, value per acre, improvement-to-land ratio, surface parking, and more —
for cities across the US and beyond.
Live site: civicmapper.org
Maintained by the Center for Land Economics.
How it works
Each city's parcel data is cleaned and exported to
GeoParquet by a Python ETL notebook or script.
Larger cities also get PMTiles vector tiles (with an H3 hexagon overlay) so
they render at any scale. The frontend loads these files over HTTP and renders
them with MapLibre GL; a small Express proxy serves the files with CORS and
rate limiting in deployed environments.
Tech stack
- Frontend — TypeScript, Vite, MapLibre GL JS, PMTiles, hyparquet/geoparquet, three.js
- API — Node.js, Express (a thin data proxy: CORS, rate limiting, blob pass-through)
- ETL — Python 3.11+, GeoPandas, Jupyter notebooks, tippecanoe/pmtiles
Repository layout
| Path | What it is |
|---|
viz/ | The map frontend (Vite + MapLibre). See viz/README.md |
api/ | Express data proxy. See api/README.md |
data/ | Per-city ETL notebooks, shared Python utilities, post-processing scripts. See data/README.md |
docs/ | Design docs and playbooks (add a city, parquet format, …) |
scripts/ | Repo-level utilities (e.g. headless smoke test of every deployed city) |
site/ | Static landing page for civicmapper.org |
.github/ | CI/CD workflows (deploy, ETL runs, smoke tests) |
Quickstart
Prerequisites: Node 20+ (frontend/API), Python 3.11+ (only needed for
the ETL side).
The one-shot way — starts the API on port 8080 and the Vite dev server on
port 5173, installing dependencies on first run:
./start-local.sh # macOS / Linux
start-local.bat # Windows (opens two terminal windows)
Or manually:
# Terminal 1 — API (optional for local dev; the frontend has its own dev proxy)
cd api
npm install
npm start # http://localhost:8080
# Terminal 2 — frontend
cd viz
npm install
npm run dev # http://localhost:5173
Open http://localhost:5173 — it takes you to the city picker
(cities.html); each city opens as app.html?city=<key>.
No environment variables are required for a basic local run: in dev mode the
Vite server proxies dataset requests to a hosted data endpoint, so cities load
out of the box. To run against your own data, either:
- drop a
<city>-<state>-parcels.parquet file into viz/public/ — in dev the
app prefers a local copy of the current city's file when one exists, or
- point
VITE_PARQUET_BASE_URL (and friends — see
viz/README.md) at your own parquet/PMTiles hosting.
For the ETL side (producing those parquet/PMTiles files yourself), follow
data/README.md.
Adding a city
The end-to-end procedure — sourcing parcel data, ETL, PMTiles, frontend
registration, deployment — is documented in
docs/add-city-playbook.md. The required output
format is specified in
docs/parcel-parquet-format.md.
Want a city but don't want to build it? Open a
city request issue.
Deploying your own instance
Civic Mapper works like a library: the engine is generic, and a deployment is
the engine plus configuration pointing at your data hosting. To stand up
your own instance:
- Host the data — parquet/PMTiles/metadata files on any static host that
supports HTTP range requests (object storage, a CDN, or the bundled
api/ proxy in front of a private bucket). Produce the
files with the ETL toolkit, following the
parquet format spec.
- Point the frontend at it — set
VITE_PARQUET_BASE_URL (and friends;
the full list with docs is in
viz/env/.env.example). ETL-side variables are in
data/.env.example. Secrets never live in this repo —
deployments supply them via environment / CI secrets.
- Register your cities — each city is a drop-in
viz/src/cities/<key>.json + viz/src/dictionaries/<key>.json; no shared
files to edit.
- CI/CD — the workflows in
.github/workflows/ are the maintainers'
deploy pipelines and are guarded by repository owner, so they skip
automatically on forks (no failing runs). Copy one as a starting point for
your own target and supply your own repository secrets.
Data attribution
Parcel and assessment data come from each jurisdiction's open-data programs.
Per-source licenses and credits are listed in ATTRIBUTION.md.
Contributing
See CONTRIBUTING.md. Security reports: see
SECURITY.md.
License
MIT.
