archive.aero ✈️

Nearly a century of U.S. aeronautical charts, scrubbable like a time machine.

archive.aero is a free, open-source web viewer for historical FAA VFR Sectional charts — 7,668 chart editions across 3,741 edition dates (as of September 2026), spanning 1930 to today, georeferenced, mosaicked, tiled, and served from edge storage. Drag the timeline and watch airspace, airports, and cartography evolve across nine decades.

⚠️ Every chart on the site is a historical scan. Never use it for navigation.

Live site: archive.aero · How it works: archive.aero/about

---

What it does

• Timeline scrubbing through every U.S. Sectional edition since 1930, with play/pause animation and keyboard shortcuts
• Seamless nationwide mosaics — individual chart sheets are stitched into a single layer per edition, so you pan the whole country at any date
• Fast on any device — static site, no application backend; tiles stream straight from object storage as WebP
• Shareable views — URLs encode date, position, and zoom
• Mobile-aware — touch controls, geolocation, and iOS Safari memory management tuned for tablets in the cockpit (for history browsing, not navigation!)

Site publication

GitHub Pages uses the Deploy site to GitHub Pages Actions workflow. It publishes only _site/, staged by python3 scripts/build_pages.py from site-files.json and committed, fingerprinted assets/{boot,viewer,csv}.*.js modules. Old module hashes stay available for cached HTML. The workflow then minifies those staged modules with esbuild (scripts/minify_site.mjs); the committed copies stay readable, and npm run test:browser:site runs the browser suite against the minified artifact. dates.csv is the viewer's public fallback inventory (date and chart URL); operational CSVs are excluded.

Worklists, scripts, source trees, local catalogs, credentials, and repository documentation are excluded from the site artifact. The GitHub repository is public: tracked files and history remain accessible there. Adding a new published file requires reviewing and updating the allowlist. Pages must use GitHub Actions as its publishing source, not a branch root.

License

Original archive.aero software and associated documentation are licensed under the MIT License, including the frontend, Workers, and processing scripts. Archived charts, scans, photographs, source datasets, and other third-party content are not relicensed by this grant; their rights and source terms remain applicable. See Sources & Attribution. Vendored libraries retain their own licenses and third-party notices.

Architecture

The system is three independent pieces: a data pipeline that turns archival scans into tile archives, edge storage/delivery, and a static frontend.

flowchart LR
    A["NARA scans<br/>(Record Group 237)<br/>+ current FAA charts"] --> B["slicer.py<br/>Python + GDAL<br/>georeference · crop collars<br/>warp to Web Mercator · mosaic"]
    B --> C["geotiff2pmtiles<br/>Go<br/>GeoTIFF → PMTiles<br/>WebP tiles, Hilbert-ordered"]
    C --> D["Cloudflare R2<br/>one .pmtiles archive<br/>per chart edition"]
    D --> E["Cloudflare Worker<br/>HTTP range proxy<br/>caching · CORS · analytics"]
    E --> F["Static frontend<br/>Leaflet + PMTiles<br/>timeline UI · tile caching"]

1. Chart processing pipeline — slicer.py (Python + GDAL)

Takes raw chart scans (from the U.S. National Archives for historical editions, FAA digital products for current ones), georeferences them, crops the paper collars, warps everything to Web Mercator, and mosaics the individual sheets into one nationwide GeoTIFF per edition date. Handles scans that are actually PDFs, mixed projections, and decades of inconsistent FAA cartographic conventions.

2. Tile conversion — geotiff2pmtiles (Go, separate repository)

A standalone, memory-efficient converter from GeoTIFF to PMTiles single-file tile archives, with native WebP encoding, multiple resampling methods, and Hilbert-curve tile ordering. It is an external CLI dependency from pspoerri/geotiff2pmtiles. The PMTiles run-length fix was merged upstream in PR #49.

slicer.py --chart-pmtiles ... checks upstream main once at startup and installs that commit with CGO_ENABLED=1, preserving lossy WebP quality 80. Binaries and build provenance are cached by commit outside the project: ~/Library/Caches/archive.aero/geotiff2pmtiles/ on macOS, or ${XDG_CACHE_HOME:-~/.cache}/archive.aero/geotiff2pmtiles/ on Linux. An existing build is reused; a missing or newer revision is downloaded and built. The selected binary stays fixed throughout the batch. A failed update stops the run rather than silently using stale code. Ordinary GeoTIFF-only slicer runs do not fetch or build g2p.

Prerequisites: Git, Python 3, Go with automatic toolchain downloads enabled (GOTOOLCHAIN=auto, Go's default), a C compiler, pkg-config, and libwebp. On macOS: brew install go pkg-config webp plus the Xcode command-line tools. Upstream's go.mod controls the required Go version.

For standalone mosaic conversion, resolve the current dependency once per batch:

g2p="$(python3 scripts/install_geotiff2pmtiles.py)" || exit 1
"$g2p" -format webp -quality 80 input.tif output.pmtiles

The installer also updates bin/geotiff2pmtiles inside that user cache for direct CLI use. To rerun a specific build or work offline, pass --geotiff2pmtiles-bin /absolute/path/to/geotiff2pmtiles to the slicer; this explicit override skips the update. A development checkout is not required; keep any checkout used for upstream contributions outside this project.

3. Delivery — worker/ (Cloudflare Worker + R2)

Each edition is one .pmtiles file in R2. Its key never changes, but an improved build is republished in place under the same key (URI-POLICY covenant 7), so the bytes behind a key can change; the next/ viewer's archives use content-hashed keys instead. A small Worker proxies HTTP range requests to R2, adds caching and CORS, and logs sampled usage and every failed request to Analytics Engine — no tile server, no database, no application backend. See worker/README.md.

4. Frontend — index.html + src/viewer.js (vanilla JS + Leaflet)

A single-page app with no framework or bundler. Native ES modules are fingerprinted by a small Node script; the generated assets are committed, so static hosting needs no build command. The interesting parts:

• Double-buffered chart layers — the outgoing edition stays visible until the incoming one is ready, so scrubbing the timeline never flashes the basemap
• LRU tile caches (decoded bitmaps + raw data) with backpressure, plus reduced limits and ImageBitmap fallbacks for iOS Safari's memory constraints
• Prefetching of adjacent editions while you scrub, with reference-counted in-flight loads
• Accessible UI — keyboard shortcuts (← → Space F S ?), ARIA labels, focus management

Frontend changes and dependencies

Edit src/, then run npm run build:frontend and commit the generated assets/ files with index.html. npm run check:frontend detects stale artifacts and checks vendored bytes against the locked npm packages and the existing SRI hashes. The metadata/airspace builders still accept --update-html index.html; they now update src/viewer.js and regenerate the module graph automatically (requires Node).

Leaflet 1.9.4 (including CSS/images), PMTiles 3.0.6, Protomaps Leaflet 5.0.0, and PapaParse 5.4.1 are served from versioned vendor/ paths. Reproduce them with npm ci && npm run vendor:frontend. PapaParse loads only when the metadata bundle cannot be used and the viewer needs dates.csv. Normal startup requests no unpkg resources. Fonts, analytics, geolocation and chart-data origins are unchanged.

Content-hashed module URLs and versioned vendor URLs are safe to cache. Keep prior deployed hashes available for cached HTML; the build does not delete them. Actual cache headers remain controlled by the existing GitHub Pages/Cloudflare hosting, not a repository _headers file. Do not edit vendored files or hashed outputs in place; see vendor/README.md for dependency update details.

5. ATC History collection — worker-atc/ (Cloudflare Worker + R2)

archive.aero is also the permanent home of the Air Traffic Control History collection — the atchistory.org archive of FAA Flight Service Station history (facility photos, training class photos, airway maps, scanned publications), flattened to a fully static site. A dedicated Worker serves it from the atc-site R2 bucket under /atc/ and 301s every old atchistory.org URL to its preserved page. The content itself (≈3.6 GB) lives only in R2 and the offline build tree — never in this repo.

The facility-location browser is restored by worker-atc/src/facilities.js: native state dropdowns link to server-rendered city galleries. Its public metadata index (303 locations, 993 entries) is generated from the preserved site with python3 scripts/atc_build_facilities.py. After refreshing that index or changing the rendered markup, increment FACILITIES_VERSION to invalidate HTML cache entries and validators, then deploy worker-atc. Original articles and photographs remain in R2.

Data sources

Historical charts are digitized scans held by the U.S. National Archives (Record Group 237 — Records of the Federal Aviation Administration); current editions come from FAA digital products. Full attribution and licensing: archive.aero/sources.

Roadmap

• Terminal Area Charts (TACs) alongside Sectionals
• Worldwide historical coverage
• Keep the site free — if hosting costs ever demand it, aviation-related sponsorships before paywalls

About

Built and maintained by Ryan Hemenway, a private pilot who wanted a "ForeFlight Time Machine" and decided to build one. If the project is useful to you, you can support it here.