English | Français

lidar2map

Offline archaeological LiDAR maps, multi-country + IGN raster/vector + OSM, for Locus Map / OsmAnd / TwoNav

A self-contained tool (standalone executables for Windows / macOS / Linux, no Python required; also runs as a single Python script) that downloads public LiDAR data from national portals across 27 countries (France, UK, Germany, Austria, Netherlands, Switzerland, Norway, Belgium, Luxembourg, Finland, Denmark, Sweden, Ireland, Czechia, Slovenia, Estonia, Latvia, Spain, Portugal, Italy, Poland, USA, Canada, New Zealand, Australia, Philippines, Japan), computes relief visualizations tuned for archaeological prospection, and generates maps usable offline on a smartphone (MBTiles, RMAP, SQLiteDB, Mapsforge formats). The IGN raster/vector maps remain France-only.

The same extent under three views: aerial imagery and the OSM map show nothing of the micro-relief, the Sky-View Factor computed from the HD LiDAR reveals it instantly.

⚠️ Status: personal project, publicly released. Heavily tested on Windows 10/11. Linux and macOS tested partially, known cases + cross-OS troubleshooting in the Troubleshooting section of BUILD.md. Feedback welcome via GitHub issues.

Note: the GUI auto-detects your language (English/French, with a manual toggle) and the CLI flags and --help are in English. The former French flag names still work as aliases, so older example commands keep working.

---

Is your country covered? 27 countries with bare-earth LiDAR (incl. USA, Canada & Japan, project-based). Check your area before diving in:

Resolutions, codes and evaluated sources: see the LiDAR coverage section below.

---

Who is it for?

• Amateur archaeologists interested in LiDAR prospection: the tool works across 27 countries (France, UK, Germany, Austria, Netherlands, Switzerland, Norway, Belgium, Luxembourg, Finland, Denmark, Sweden, Ireland, Czechia, Slovenia, Estonia, Latvia, Spain, Portugal, Italy, Poland, USA, Canada, New Zealand, Australia, Philippines, Japan) with more in progress. The relief computations (multi, SVF, openness, LRM, RRIM, VAT) are identical from one country to the next.
• French hikers who want offline IGN topo maps on their phone (Locus Map Pro, OsmAnd+): the IGN raster/vector tabs remain France-only.
• Landscape surveyors who combine historical orthophotos (1950-1995, France) with a DEM to spot human remains before agricultural land abandonment erases them.
• Cavers / explorers who need accurate base maps in areas not covered by mainstream apps.

The tool is not intended for metal detecting. The code strictly respects the open licenses involved (Etalab FR, CC BY 4.0 NO, CC-0 NL, BGDI CH).

What it produces

From a town, GPS coordinates, a bbox, a département or a whole region:

• Archaeological relief from national LiDAR (0.5 m to 1 m resolution depending on source):

  Type	What it reveals	Parameters
  multi	Multidirectional hillshade (Mark 1992), general relief without azimuth bias	elevation (° sun, default 25, low = micro-relief, 45 = general use)
  315 045 135 225	Directional hillshades, emphasize structures perpendicular to the chosen azimuth	elevation (same)
  slope	Slope 0-90° stretched to 1-255, banks, breaks, terraces	(none)
  svf	Sky-View Factor, fraction of visible sky: ditches, terraces, enclosures shown dark	conv (flux = cos²γ contrasted, default; rvt = 1−sin γ, the Kokalj/Hesse archaeology standard), dist (horizon radius in m, default 20, 20 = micro-relief, 100 = enclosures/roads), gamma (contrast, default 2.0)
  opos	Positive openness (Yokoyama 2002), mean horizon angle above the horizontal: ridges, mounds, barrows shown bright	dist, gamma
  oneg	Inverted negative openness, the "looking down" view: ditches, banks and hollow ways shown dark, the SVF's companion (inherently grainier: sensitive to DTM noise)	dist, gamma (applied mirrored: deepens hollows without darkening the background)
  lrm	Local Relief Model, subtracts the smoothed terrain (gaussian σ): removes hills and valleys, keeps only local anomalies. Fast and readable: the GUI default	sigma (gaussian radius in m ≈ max scale of preserved structures; default 15 px of the provider)
  rrim	Red Relief Image Map (Chiba 2008), color composite: slope in red (absolute 0-45° ramp), LRM as light/dark, hollows AND mounds at a glance	sigma (of the internal LRM)
  vat	Visualization for Archaeological Topography, the most complete detector: SVF + positive openness + slope blended into a single grayscale, reveals hollows AND mounds without picking a method (RVT style, ZRC SAZU). Slower than lrm, grainier too. Needs numba	dist (SVF/openness radius in m, default 20), gamma (final composite contrast, default 2.0, 1 light, 2 dark)

  Two ways to request them:

  # Simple: list of types, shared global parameters
  --shadings multi svf oneg --svf-dist 20 --svf-gamma 2
  
  # Parameterized instances (repeatable): each occurrence carries ITS OWN params
  # → several instances of the same type in a single run
  --shading svf:dist=20,gamma=2 --shading svf:dist=100,gamma=1.5 \
  --shading oneg:dist=20 --shading 315:elevation=20 --shading lrm:sigma=10
  
  # Resolution preset (opt-in): a stack (svf + opos + lrm + multi + slope) sized
  # in METRES for the DEM resolution, so the same ground-scale features are
  # targeted whether the DEM is 0.25 m or 5 m. 'auto' picks the tier per provider:
  #   micro (<=0.75 m) / standard (~1 m) / landscape (>=5 m)
  --shading-preset auto

  Explicit parameters that differ from the defaults are encoded in the output filename (zone_svf_flux_100m_g1p5_ombrage.tif, zone_315_e20_ombrage.tif): no collision between instances, and already-computed shadings are reused. In the GUI, the "to process" list (+/− buttons) does the same: each added instance has its own little parameter form. --svf-sweep / --no-svf-sweep (sweep-horizon kernel, SVF only) stays global.

  Known limit — standing ruins. National bare-earth DTMs remove by design walls still standing above ~1 m: the classifier files them as vegetation or "unclassified" (the IGN spec documents this for roofless ruined buildings), and the DTM interpolates straight through them. Typically (observed, not a guaranteed rule), a 40 cm enclosure wall survives (absorbed into the ground class) while a 1.5 m house ruin vanishes cleanly. No shading computed from the DTM can bring them back. For targeted prospection of standing structures in France, use tools/dfm_ruines.py: it rebuilds a DFM-style model (the Digital Feature Model concept is from Štular et al. 2021; the automatic point selection used here is a first-pass heuristic — the literature does this step by (semi-)manual reclassification) from the classified IGN LiDAR HD point cloud (COPC LAZ, ~205 MB/km²) by re-injecting low non-ground returns (0.4-2.5 m) into the ground-class gaps, and outputs georeferenced LRM-DTM / LRM-DFM / delta GeoTIFFs to drape over the orthophoto in QGIS. Walls show up as thin continuous lines; scrub shows as speckle — the eye does the final discrimination. The same DFM is also built into the pipeline: tick the "DFM mode" checkbox next to the provider (or CLI --dfm) and all shadings (LRM, VAT…) run on the DFM instead of the DTM, at the point-cloud download cost — keep the area small. Height band and LAS classes are tunable per site (GUI fields / --dfm-hmin, --dfm-hmax, --dfm-classes); the LAZ stays cached so retuning re-converts in ~20 s without re-downloading. Alternative ground base: --dfm-ground csf (GUI "ground base" select) replaces the class-based re-injection with a Cloth Simulation Filter (Zhang et al. 2016): a soft simulated cloth absorbs low continuous structures into the ground while rejecting vegetation, entirely ignoring the producer's classes. Cleaner background (no speckle), same wall signal on the test sites; ~3 min/tile instead of ~20 s. The cloth is tunable per site with the standard CSF surface (--dfm-csf-threshold, --dfm-csf-resolution, --dfm-csf-rigidness 1 steep / 2 / 3 flat; same fields in the GUI). DFM mode is not France-only: it also runs on Switzerland's swissSURFACE3D point cloud (--provider ch-swisstopo --dfm, CSF ground base by default). Any provider that publishes a full, dense, classified point cloud can get a DFM twin; a bare-earth DTM raster or a ground-only cloud cannot.

  A roofless house ruin (walls ~1.5 m, dép. 83, France), under scrub. The aerial photo barely hints at the walls; the classic LRM (from the DTM) shows the terraces but not the ruin; the DFM brings the building footprint back, and the CSF ground base cleans up the background.

  Aerial ortho	Classic LRM (from the DTM)
  Walls lost under vegetation	Terraces show, the ruin does not
  DFM-LRM (class re-injection)	DFM-LRM (CSF cloth base)
  Rectangular building reappears (speckly)	Same walls, cleaner background

  LiDAR sources: 27 countries via the --provider <code> flag (or the GUI dropdown), France (default), Netherlands, Switzerland, Norway, Germany (12 Länder), Austria (national + Tyrol), United Kingdom, Belgium (Flanders), Finland, Denmark, Ireland, Czechia, Slovenia, Estonia, Latvia, Spain (+ Basque Country, Navarre, Catalonia), Italy (Emilia-Romagna, Sardinia), Poland, USA, Canada, New Zealand, Australia (QLD/NSW), Philippines (Taal area). Per-provider details (dataset, resolution, CRS, access mechanism, coverage, API keys) live in the single reference table of the LiDAR providers section.

• IGN raster maps (France only): Plan IGN, orthophotos (current + historical 1950, 1965, 1980), 19th-century État-Major, Pléiades satellite, CIR, etc.

• USGS Imagery (USA, --layer naip): public-domain NAIP-derived aerial imagery (~1 m, cache complete to z16), pairs with the 3DEP LiDAR (us-tnm).

• Vector maps: OSM Mapsforge .map (international, via Geofabrik) or IGN BD TOPO (France only). Both can also render as transparent-raster: the selected layers (paths, roads, rivers...) drawn on transparent tiles (.sqlitedb), to float above the LiDAR relief as an OsmAnd overlay (OsmAnd cannot overlay vector data natively)

• Outputs: MBTiles (universal), RMAP (CompeGPS / TwoNav), SQLiteDB (RMaps schema, Locus Map / OsmAnd), Mapsforge .map (Locus Map), transparent overlay .sqlitedb (transparent-raster)

• Send to phone: after generating, the GUI's 📲 button (or --serve --zone-name X in CLI) serves the maps on your local WiFi and shows a QR code. Scan with the phone, download, then "Open with" OsmAnd or Locus: no cable, no cloud, nothing leaves your network. (Android may warn the download is insecure: choose Save, it is a plain local transfer.)

• Processing queue: in the GUI, stack several zones with the ＋ Queue button, then Run queue runs them one after another, unattended. A failed job doesn't stop the queue (each item shows its status), so you can line up a batch of areas and walk away. The CLI equivalent is chaining commands in a shell script.

• Index sheet: each run drops a <product>_planche.png next to the deliverables, showing the coverage extent, the real department outline (with a locator inset when the view is zoomed in), and the numbered chunk cells when the area was split. One sheet per map product (each shading gets its own); the vector layers of a run share a single sheet. Built by scanning the actual files (mbtiles/sqlitedb/geojson), so you can also regenerate it for any existing project folder with --index-sheet DIR, without re-running anything. Disable per-run with --no-index-map.

---

Installation and usage

Quick start: download the standalone executable for your OS from the Releases page, extract, run. No Python, no dependencies, nothing to install.

Two ways to use lidar2map:

	A. Standalone executable	B. Python script
Requirements	None	Python 3.12
First install	None	~5 min (auto bootstrap in its own venv)
Updates	Patch the 3 existing binaries on the GitHub release in one command: python update_app.py --release (see update_app.py)	git pull + relaunch
Distributable	Yes, .exe / .app / Linux binary + zip bundle side by side	No, each user installs Python
Best for	end user / Windows / distributing	dev / Linux / contributing code

A. Standalone executable

No Python for the end user to install. The deliverable carries its own runtime (embedded Python, deps, JRE, osmosis).

1. Get the deliverable

Option a, Download from Releases (if the version is published for your platform):

OS	Archive	Extract with
Windows 10/11 (x86_64)	lidar2map-windows-x86_64.zip	Expand-Archive (PowerShell) or double-click
Linux Ubuntu 24.04+ (x86_64)	lidar2map-linux-x86_64.tar.gz	tar xzf
macOS 12+ (Apple Silicon)	lidar2map-macos-arm64.zip	unzip then xattr -dr com.apple.quarantine LIDAR2MAP.app

The archive extracts into a lidar2map-<os>-x86_64/ folder containing the binary and its lidar2map_bundle.zip side by side. No system installation.

Option b, Build it yourself. Two scripts per platform: a machine setup (do once) then a build (re-run each time lidar2map.py is updated).

Windows

git clone https://github.com/nico579/lidar2map
cd lidar2map
.\setup_build_windows.ps1     # 1. Setup: Python 3.12, deps, JRE, osmosis, PyInstaller
.\lidar2map_win_build.ps1     # 2. Build: 3 steps -> dist\lidar2map.exe + dist\lidar2map_bundle.zip

macOS (Apple Silicon)

git clone https://github.com/nico579/lidar2map
cd lidar2map
bash setup_build_mac.sh       # 1. Setup
bash lidar2map_mac_build.sh   # 2. Build -> dist/LIDAR2MAP.app

Linux (Ubuntu / Debian)

Linux reuses the Windows specs (_win.spec produces an ELF on Linux, the name is misleading).

git clone https://github.com/nico579/lidar2map
cd lidar2map
bash setup_build_linux.sh       # 1. Setup
bash lidar2map_linux_build.sh   # 2. Build -> dist/lidar2map + dist/lidar2map_bundle.zip

Requirement: sudo apt install zip if missing. The produced binary depends on the build machine's libc (build on Ubuntu 22.04 → runs on Ubuntu ≥ 22.04 / Debian 12+).

Full build documentation (bundle architecture, updating without rebuild, troubleshooting): BUILD.md.

2. Run the deliverable

OS	Command
Windows	Double-click lidar2map.exe (or run from a terminal to see the log)
Linux	chmod +x lidar2map && ./lidar2map in the extracted folder
macOS	Double-click LIDAR2MAP.app. First launch blocked by Gatekeeper: xattr -dr com.apple.quarantine LIDAR2MAP.app then double-click
Linux	chmod +x lidar2map && ./lidar2map

The first launch extracts the bundle (~30-60 s, once, it contains Qt) into:

• Windows: %LOCALAPPDATA%\lidar2map\
• macOS: ~/Library/Application Support/lidar2map/
• Linux: ~/.local/share/lidar2map/

Clean uninstall: lidar2map(.exe) --desinstaller.

B. Python script

On first launch, the script creates ~/.lidar2map/venv and installs the critical dependencies there (Pillow, pyproj, numpy, rasterio, pywebview + PyQt6/QtWebEngine…): your system Python is never touched (--bootstrap=none if you prefer to manage the environment yourself). The Temurin 21 JRE and osmosis are downloaded on demand; no system GDAL needed, the rasterio wheels embed their own. ~400 MB total, once.

Windows 10+

1. Install Python 3.12+
2. Get the code:

   git clone https://github.com/nico579/lidar2map
   cd lidar2map
   python lidar2map.py

macOS 11+

brew install python@3.12
git clone https://github.com/nico579/lidar2map
cd lidar2map
python3.12 lidar2map.py

Linux (Debian / Ubuntu)

sudo apt install python3.12 python3.12-venv git
git clone https://github.com/nico579/lidar2map
cd lidar2map
python3.12 lidar2map.py

Troubleshooting: the Troubleshooting section of BUILD.md (including Linux/macOS-specific cases: PEP 668, Qt distro packages, Wayland, Gatekeeper on the JRE…).

---

Usage

Two modes, selected automatically based on the arguments (same logic as the twin project gpxsolar):

• No argument → graphical interface (pywebview / Qt). The common mode.
• With arguments → command-line computation (headless, no window). Handy for scripting, running on a server, or reproducing an exact render.

Everything below applies to the binary as well as the script, just replace python lidar2map.py with lidar2map.exe (Windows), ./lidar2map (Linux) or LIDAR2MAP.app (macOS).

Command-line examples

The flags below are English. The former French flag names still work as aliases, so older commands keep working.

SVF relief + IGN topo map over a town (1 km² zone around Garéoult, France):

python lidar2map.py --lidar --zone-city Gareoult --zone-radius 1 \
    --shadings multi svf --file-formats mbtiles```

**Relief over Amsterdam (Netherlands, AHN4):**
```bash
python lidar2map.py --provider nl-ahn --lidar --download \
    --zone-bbox 120000,486000,122000,488000 --zone-name amsterdam \
    --shadings multi --file-formats mbtiles```

**Relief over Geneva (Switzerland, swissALTI3D):**
```bash
python lidar2map.py --provider ch-swisstopo --lidar --download \
    --zone-city Geneve --zone-radius 1 \
    --shadings svf --file-formats mbtiles```

**Relief over Oslo (Norway, Kartverket):**
```bash
python lidar2map.py --provider no-kartverket --lidar --download \
    --zone-city Oslo --zone-radius 1 \
    --shadings multi --file-formats mbtiles```

**Historical 1950-1965 orthophoto over an archaeological survey area:**
```bash
python lidar2map.py --raster --zone-bbox 6.0,43.3,6.1,43.4 \
    --layer ortho_1950 --zoom-min 14 --zoom-max 18```

**OSM vector map (Mapsforge .map) for Locus, whole département:**
```bash
python lidar2map.py --osm --zone-department 83 --file-formats map```

**Whole region (`--zone-region`), available for all modes:**
```bash
# OSM: a single map for the whole region, no re-splitting
# (the Geofabrik PBF IS already regional, far faster than looping per département)
python lidar2map.py --osm --zone-region provence-alpes-cote-d-azur
# IGN vector: paths/routes for the whole region as GeoJSON + Locus .map
python lidar2map.py --vector --zone-region provence-alpes-cote-d-azur \
    --layer chemins --file-formats gz map```
The slug is the one from [Geofabrik France](https://download.geofabrik.de/europe/france.html) (old-style regions: `provence-alpes-cote-d-azur`, `bretagne`, `corse`, `rhone-alpes`…). In OSM the region is processed as one block (the Geofabrik file is already regional, no per-département geocoding); for the raster/vector/lidar modes the area is the bbox enclosing all the départements of the region. An unknown slug lists the available regions.

**IGN BD TOPO map (roads + buildings) as compressed GeoJSON + Mapsforge .map:**
```bash
python lidar2map.py --vector --zone-department 83 \
    --layer routes batiments --file-formats gz map```
The `map` format converts the IGN GeoJSON into a Mapsforge `.map` map (readable by Locus Map; OsmAnd uses its own OBF vector format and cannot read Mapsforge files, but its built-in offline map already provides the vector layer, so on OsmAnd simply put the LiDAR raster on top as an overlay).

## LiDAR providers, adding a country

The provider abstraction lets you add a national LiDAR source without touching the core of the pipeline. Each provider lives in `providers/<code>.py` (~50-200 lines) and exposes:

```python
NAME, CODE, COUNTRY, LICENSE          # metadata
CRS_NATIF, RESOLUTION_M, DALLE_KM     # geometry
discover_dalles(bbox_wgs, bbox_natif, cache)  # → {name: url}
# + helpers: dalle_filename, dalle_url, subdir_from_name, dalles_pour_bbox

The downstream pipeline (SVF, relief, EPSG:3857 warp, MBTiles) is provider-agnostic: it consumes the GeoTIFFs returned by discover_dalles, regardless of the native CRS or the index format used upstream.

Code	Country	Dataset	Res.	Native CRS	Access & specifics
fr-ign	France (default)	IGN LiDAR HD	0.5 m	EPSG:2154 (Lambert-93)	Vector TMS PBF + WMS GetMap, national coverage (mainland)
fr-reunion · fr-guadeloupe	France (Réunion, Guadeloupe DROM)	IGN LiDAR HD	0.5 m	EPSG:2975 / 5490 (UTM40S / UTM20N)	WFS IGNF_MNT-LIDAR-HD:dalle index (each tile feature carries its direct download url), 0.5 m GeoTIFF, Licence Ouverte 2.0 (Martinique/Mayotte announced but WFS empty for now)
fr-ign + DFM mode	France (standing-ruins mode, experimental)	DFM from classified LiDAR HD point cloud	0.5 m	EPSG:2154 (Lambert-93)	GUI checkbox "DFM mode" (or CLI --dfm, with --dfm-hmin/--dfm-hmax/--dfm-classes to tune per site): downloads the COPC LAZ tiles (~205 MB/km²!) and rebuilds the model from ONE class set (default 1,2,3,4,9,66: classes 2/9/66 = terrain base as in the official DTM, the others are re-injected into ground gaps within the 0.4-2.5 m height band). Can re-introduce returns compatible with standing walls that the DTM erases (candidates, not a wall classifier — scrub comes back too; see "Known limit" box). Alternative ground base --dfm-ground csf (Cloth Simulation Filter, Zhang et al. 2016): ignores the producer's classes entirely, cleaner background, ~3 min/tile; cloth tunable per site (--dfm-csf-threshold/-resolution/-rigidness, standard CSF surface). (Removing class 2 from the set yields a slice, band objects only on a transparent background; rarely useful in practice.) The zone name is auto-suffixed (_laz_dfm / _laz_csf: laz = the point-cloud source, dfm/csf = the method; the DTM default stays unmarked), so point-cloud outputs land in their own project and never mix with DTM ones. The LAZ is kept in the tile cache: changing the settings re-converts without re-downloading. Targeted prospection of a few km², not large maps
nl-ahn	Netherlands	AHN4/5	0.5 m	EPSG:28992 (RD New)	ATOM feed + JSON FeatureCollection, national coverage
ch-swisstopo	Switzerland	swissALTI3D	0.5 m	EPSG:2056 (CH1903+/LV95)	STAC REST API, national coverage
ch-swisstopo + DFM mode	Switzerland (standing-structures mode, experimental)	DFM from classified swissSURFACE3D point cloud	0.5 m	EPSG:2056 (CH1903+/LV95)	GUI checkbox "DFM mode" (or CLI --dfm) on the Swiss provider: downloads the swissSURFACE3D .las.zip tiles (~125 MB/km²) via the same STAC API, unzips the point cloud and rebuilds the standing-structures model. Default ground base is CSF (--dfm-ground csf, Cloth Simulation Filter) since swisstopo's class codes are not guaranteed IGN-compatible; the classes mode is also available. Same per-site tuning and cache-then-retune behaviour as the France DFM (~6 min/tile). Targeted prospection, field validation recommended
no-kartverket	Norway	Nasjonal Høydemodell	1 m	EPSG:25833 (UTM33N)	ArcGIS ImageServer exportImage, national coverage
se-lantmateriet	Sweden	Markhöjdmodell (laser)	1 m	EPSG:3006 (SWEREF99 TM)	STAC + 10 km mosaic COG (windowed read), national coverage; free GeoTorget account (env LANTMATERIET_USER/LANTMATERIET_PASS) for the download
de-bayern · de-nrw · de-niedersachsen · de-rlp	Germany (4 Länder: Bavaria, NRW, Lower Saxony, Rhineland-Palatinate)	DGM1	1 m	EPSG:25832 (UTM32N)	metalink / index.json / STAC COG, open data (de-rlp: Metalink index of ~21k GeoTIFF tiles, post_fetch strips the compound vertical CRS to 25832)
de-thueringen · de-berlin · de-sh	Germany (Thuringia, Berlin, Schleswig-Holstein)	DGM / DGM1	1-2 m / 1 m	EPSG:25832 / 25833 (UTM32N/33N)	Spatial index (ATOM or GeoJSON) → XYZ text tiles (post_fetch → GeoTIFF), open data (Thuringia/SH CC BY / dl-de/by-2-0, Berlin dl-de/zero-2-0)
de-hessen · de-bw · de-mv · de-st · de-brandenburg	Germany (Hesse, Baden-Württemberg, Mecklenburg-Vorpommern, Saxony-Anhalt, Brandenburg)	DGM1	1 m	EPSG:25832/25833 (UTM32N/33N)	WCS 2.0.1 INSPIRE GetCoverage, open data dl-de/by-2-0 (de-mv/de-st found via the GDI-DE catalog auto-discovery)
at-bev	Austria (national)	ALS-DGM	1 m	EPSG:3035 (LAEA Europe)	ATOM index + 50 km mosaic COG (windowed read via /vsicurl), latest survey per tile, CC BY 4.0 (BEV)
at-tirol · at-osttirol	Austria (Tyrol + East Tyrol)	DGM	0.5 m	EPSG:31254/31255 (MGI M28/M31)	WCS 1.0.0 GetCoverage (tiris), finer than at-bev over Tyrol
gb-england · gb-wales	United Kingdom	LIDAR Composite DTM	1 m	EPSG:27700 (OSGB36)	WCS 2.0.1 / WFS catalogue (EA / NRW)
gb-scotland	United Kingdom (Scotland)	Scottish Public Sector LiDAR DTM	0.5 m	EPSG:27700 (OSGB36)	Public AWS S3 bucket (no account), OS-grid tile listing (ListObjectsV2) → COG, modern 50 cm coverage (national programme + Orkney)
be-flanders	Belgium (Flanders + Brussels)	DHMV II DTM	1 m	EPSG:31370 (Lambert 1972)	WCS 2.0.1, also exposes pre-computed 25 cm SVF and multi-hillshade
lu-act	Luxembourg	BD-L-Lidar 2024 DTM	0.5 m	EPSG:2169 (LUREF)	Single national COG (~40 GB) read windowed via /vsicurl HTTP range, never downloads the whole file; CC0
fi-maanmittauslaitos	Finland	Elevation Model	2 m	EPSG:3067 (TM35FIN)	WCS 2.0.1, free API key required, national coverage
dk-datafordeler	Denmark	DHM DTM	0.4 m	EPSG:25832 (UTM32N)	WCS 1.0.0, free API key required, national coverage
ie-gsi	Ireland	LiDAR DTM	1 m	EPSG:2157 (ITM)	ArcGIS FeatureServer → ZIP (post_fetch), ~60% coverage, CC BY 4.0
cz-cuzk	Czechia	DMR 5G	1 m	EPSG:5514 (S-JTSK/Krovak)	Atom INSPIRE 2-level → LAZ (post_fetch, requires lazrs), national coverage
si-arso	Slovenia	DMR1 (2011-2015 LiDAR)	1 m	EPSG:3794 (D96/TM)	ArcGIS REST fishnet index + x;y;z text tiles → GeoTIFF (post_fetch), national coverage
ee-maaamet	Estonia	DTM 1 m (2021-2024 ALS)	1 m	EPSG:3301 (L-EST97)	Direct per-sheet URLs, 1:10000 grid (sheet numbering = pure formula, no index), national coverage, open data
lv-lgia	Latvia	DTM 1 m (LiDAR ALS)	1 m	EPSG:3059 (LKS-92/TM)	S3 index of ~66k classified LAS tiles → download → class-2 binning to GeoTIFF with hole-fill (requires laspy), national coverage, CC BY 4.0 (tile extents measured from LAS headers, TKS-93 sheet grid)
es-cnig	Spain	MDT	5 m	EPSG:25830 (UTM30N)	WCS 2.0.1 INSPIRE, 5 m = landscape scale (the 2 m bare-earth LiDAR requires the session-based CNIG portal)
es-icgc	Spain (Catalonia)	MET LiDAR	0.5 m	EPSG:25831 (UTM31N)	Single regional COG (~433 GB) read windowed via /vsicurl HTTP range, 50 cm, far finer than es-cnig 5 m; CC BY 4.0 (ICGC)
es-euskadi	Spain (Basque Country)	MDT LiDAR	1 m	EPSG:25830 (UTM30N)	WCS 1.0.0 (ArcGIS MapServer WCSServer, geoEuskadi), 1 m bare-earth, far finer than es-cnig 5 m; CC BY 4.0
es-navarra	Spain (Navarre)	MDT LiDAR	2 m	EPSG:25830 (UTM30N)	WCS 2.0.1 INSPIRE (IDENA), 2 m bare-earth, NoData 3.4e38; CC BY 4.0
pt-dgt	Portugal	MDT LiDAR (2024)	0.5 m	EPSG:3763 (PT-TM06)	OGC-API + POST /search (CQL2), national coverage; free DGT account (env DGT_USER/DGT_PASS) for the authenticated download
it-emilia-romagna	Italy (Emilia-Romagna)	DTM (RER)	5 m	EPSG:7791 (RDN2008/UTM32N)	WCS 2.0.1 GetCoverage, regional coverage, CC BY 4.0 (the 0.5 m LiDAR 2023/24 is served once its coverage completes)
it-sardegna	Italy (Sardinia)	DTM (RAS)	1 m	EPSG:7791 (RDN2008/UTM32N)	WCS 2.0.1 GetCoverage (GeoServer), island-wide LiDAR mosaic with gaps (coast, towns, Gallura, river bands), clean nodata off-coverage, CC BY 4.0
it-piemonte	Italy (Piedmont)	DTM (ICE LiDAR)	5 m	EPSG:32632 (UTM32N)	WCS 1.0.0 GetCoverage (MapServer), format=image/tiff for the real Float32 (GTiff returns quantised UInt8), NoData -99, CC BY 4.0
pl-gugik	Poland	NMT (ISOK project)	1 m	EPSG:2180 (PUWG 1992)	WCS 2.0.1, open data, national coverage
ca-nrcan	Canada	HRDEM Mosaic	1 m	EPSG:3979 (LCC Canada)	STAC + mosaic COG (windowed read), ~95% of population
us-tnm · us-3dep	USA	3DEP	1 m	EPSG:3857	TNMAccess direct S3 (no account) / OpenTopography (free key)
us-cnmi	Northern Mariana Islands (US territory)	Topobathy DEM	1 m	EPSG:8693 (NAD83(MA11)/UTM55N)	Single NOAA mosaic VRT read windowed via /vsicurl (bucket noaa-nos-coastal-lidar-pds), ground-class bare earth on land + bathymetry offshore, public domain (pattern for a generic NOAA provider)
jp-gsi	Japan (partial)	DEM5A (GSI 標高タイル)	5 m	EPSG:3857	Open elevation XYZ text tiles, no account (post_fetch → GeoTIFF), partial 5 m coverage (rivers/plains/populated)
ph-taal	Philippines (Taal volcano area only)	DTM 1 m (UP TCAGP)	1 m	EPSG:32651 (UTM51N)	Static GeoJSON tile grid → direct GeoTIFF on S3 (<GRIDREF>_DTM.tif), ~20 km around Taal volcano, open data
nz-linz	New Zealand	National seamless DEM	1 m	EPSG:2193 (NZTM2000)	LINZ S3 STAC + COG (windowed read)
au-qld · au-nsw	Australia (QLD 0.5 m · NSW 5 m)	LiDAR DEM	0.5-5 m	EPSG:3857	ArcGIS ImageServer (ELVIS), per-state coverage
au-ga	Australia (national, scattered)	DEM derived from LiDAR	5 m	EPSG:3857 (served as 4283)	WCS 1.0.0 GetCoverage (Geoscience Australia) → reprojected on download, ~245,000 km² across all states (coastal + Murray-Darling), opens SA/VIC/TAS/WA beyond QLD·NSW

Selection: --provider <code> flag (CLI), LIDAR2MAP_PROVIDER env var, or the dropdown at the top of the GUI. This table is the single reference list of providers, the features section links here instead of duplicating it.

To add a new country: copy the provider closest in paradigm (WCS, STAC, ArcGIS ImageServer, direct COG, FeatureServer catalogue…) and adapt URLs/CRS/naming format. The first provider for a new paradigm takes ~½ day; subsequent ones with the same pattern take ~1-2 h. The provider roadmap documents every evaluated source, integrated and set-aside, with the precise reason and a paradigm-by-paradigm cheat sheet.

Main features

• Auto-bootstrap: no pre-installed dependency required. The script downloads on demand: Python deps (Pillow, pyproj, numpy, scipy, rasterio, whose wheels embed their own GDAL), Temurin 21 JRE, osmosis, mapwriter.
• Memory streaming: département-scale processing without saturating RAM (ijson, rasterio windowed reads, tile-by-tile MBTiles generation).
• Clean cancellation: Ctrl+C once → stops after the current chunk. Ctrl+C twice → immediate stop.
• Resume after interruption: the same command resumes where it stopped, via a .json manifest that tracks completed chunks.
• Up-front splitting: for large areas, split into an N×N grid or ~K km squares (--split-radius, bounded chunk size, recommended at national scale), useful so you don't have to regenerate the whole area if something crashes. Per-chunk disk cleanup (--cleanup) and a free-space guard (--min-free-gb) for very large coverage.
• Crash-safe history: each run is recorded at startup (status "running") then finalized to "ok" or "ko". A hard crash (kill -9, power loss) leaves the entry visible in the UI, the trace is kept for debugging.
• Multi-provider LiDAR: a providers/<code>.py abstraction that lets you plug in any LiDAR source. Shipped providers: FR (IGN), NL (AHN), CH (swisstopo), NO (Kartverket), DE (Bavaria, NRW, Lower Saxony), AT (Tyrol, East Tyrol), GB (England, Wales), BE (Flanders WCS), FI (NLS WCS), DK (Datafordeler WCS), IE (GSI catalogue), CA (NRCan STAC), NZ (LINZ S3), AU (Geoscience Australia WCS), US (3DEP 1m, no account), covering varied API paradigms (TMS PBF, JSON FeatureCollection, STAC, ArcGIS FeatureServer/ImageServer, Metalink/index.json, per-tile WCS GetCoverage, S3 public COG). Providers can also expose pre-computed shadings (PROVIDES_SHADINGS), the pipeline downloads them directly instead of computing from the DEM (e.g. BE Flanders SVF 25 cm, multi-hillshade 25 cm). Adding a country = ~100-150 lines (see LiDAR coverage & evaluated sources below).
• Interactive GUI: 6 tabs (LiDAR, IGN raster, IGN vector, OSM, Merge, Splitting), provider selector at the top of the form (IGN Raster/Vector tabs hidden automatically for non-FR providers), history of the last 50 commands with status badges, parameter validation, live log, error modal, and a processing queue (＋ Queue) to run several zones back-to-back.
• Historical orthophoto maps: a unique combo for archaeology, SVF 2024 (current LiDAR) + 1950 ortho (before land abandonment) → reveals structures still legible 70 years later.

LiDAR coverage & evaluated sources

The colour map is at the top of the README. Interactive version (click = NAME + code):

🗺️ Interactive coverage map, rendered directly by GitHub, or droppable into geojson.io / QGIS to test a point.

Countries on the map (national bare-earth LiDAR): France · Netherlands · Switzerland · Norway · Germany (Bavaria · NRW · Lower Saxony · Thuringia) · Austria (Tyrol) · United Kingdom (England · Wales · Scotland) · Belgium (Flanders) · Luxembourg · Finland · Denmark · Ireland · Czechia · Spain (5 m; Catalonia 0.5 m) · Poland · New Zealand · Australia (Queensland 0.5 m · NSW 5 m · national 5 m GA, scattered). Resolutions 0.5-1 m unless noted, see the provider list above for codes and details.

The map is regenerated by coverage_map.py, which reads zone titles from providers/*.py, so the map and the GUI can't drift. Clicking a zone in the interactive GeoJSON shows its NAME and code(s).

🇺🇸 USA & 🇨🇦 Canada, supported and working, just not drawn. us-tnm / us-3dep (3DEP 1 m) and ca-nrcan (HRDEM 1 m) are fully functional, but their coverage is project/population-based (not wall-to-wall national), so a full-country polygon would over-claim, hence the note rather than a shape. Check your US area on the TNM Downloader. The USGS 1 m tiles are 10×10 km COGs, read windowed to your bbox via /vsicurl/, no full-tile download.

🇧🇪 Belgium (Flanders): a bonus, the WCS also exposes DHMV_II_SVF_25cm (Sky-View Factor at 25 cm, 16 directions, r=2.5 m) and DHMV_II_HILL_25cm (multidirectional hillshade at 25 cm, pre-computed by Digitaal Vlaanderen). When one of those shadings is requested, lidar2map downloads it directly instead of computing it from the 1 m DEM, both faster and at higher resolution.

A source plugs in cleanly when it exposes deterministic tiles (one URL per ~1 km tile), a WCS (GetCoverage by bbox), mosaic COGs (windowed /vsicurl/ read on the bbox, see ca-nrcan) or LAZ/ZIP tiles (post_fetch hook: unzip + point-cloud→GeoTIFF via laspy+lazrs, see cz-cuzk, ie-gsi). Still a poor fit: sources via form/email order, WMS only (rendered, no raw elevation) or ASC without a CRS.

Not covered yet, and why: the full registry of evaluated-but-not-integrated sources (Wallonia, Saxony, Slovakia, Northern Ireland, Latvia, Hong Kong, Taiwan, Iceland, national Italy, national Germany, and more), each with the precise blocking reason and a re-check date, lives in the provider roadmap. Kept as a single file to avoid re-digging dead ends. | Africa · rest of Asia | ⛔ structural | no open national bare-earth LiDAR (global 30 m DEMs only). | | OpenTopography (global) | ⛔ structural | fine LiDAR = point cloud / async jobs; its simple raster API is 30 m satellite. |

🔄 pending = open data but no per-bbox programmatic access yet, re-checked periodically (next review ~Dec 2026). ⛔ structural = blocked for now (data nonexistent, paid, classified, too coarse, or not bare-earth LiDAR).

Live in one of these places? You may know a way in. Most 🔄 cases just need a documented endpoint accessible by bounding box, a WCS GetCoverage, an INSPIRE ATOM feed, STAC, derivable per-tile URLs, or a public S3 bucket. If you know one for your country/region, open an issue or PR, adding a provider is ~100-150 lines (copy the closest providers/*.py). Germany is in as far as cleanly possible (4 states: Bavaria, NRW, Lower Saxony, Thuringia).

Screenshots

Graphical interface

Six tabs to drive LiDAR, IGN raster/vector, OSM, merge and splitting.

HD LiDAR (archaeological relief)	IGN raster (Plan / ortho / historical)	IGN vector (BD TOPO)

OSM vector (Mapsforge)	Vector merge	Raster splitting

Send to phone: the 📲 button serves the generated maps over local WiFi, scan the QR code and "Open with" OsmAnd or Locus.

The index sheet dropped next to the deliverables: real department outline and numbered chunk cells (here a Var department VAT run split into 3×4 zones; the slight overlaps are the real shared edge tiles at low zooms).

Rendering in Locus Map

Archaeological LiDAR relief shown as an overlay on the terrain in Locus Map.

SVF (Sky-View Factor)	Multi-hillshade overlay

Rendering in OsmAnd

LiDAR relief (LRM) as a semi-transparent Overlay map above the standard OsmAnd map (Configure map > Overlay map, transparency slider around the middle).

What SVF reveals, same area, three sources

Under tree cover, the aerial photo and OSM show nothing. The LiDAR SVF makes the terraces (dry-stone restanques) and old paths appear, invisible from above.

Satellite photo	OSM	SVF (HD LiDAR)
Opaque scrubland	Almost no detail	Crisp terraces + paths

Reproducing this render

The header SVF and the triptych above (Rougiers area, dép. 83, France) were computed with:

python lidar2map.py \
  --zone-gps <lat> <lon> --zone-radius 1 --zone-name hero \
  --lidar --download --workers 8 \
  --shadings svf --shading-elevation 25 \
  --svf-conv rvt --svf-dist 20 --svf-gamma 0.8 --svf-sweep \
  --file-formats mbtiles --zoom-min 8 --zoom-max 18 \
  --image-format jpeg --image-quality 85```

Replace `<lat> <lon>` with your own area; the SVF parameters above are the ones
used for the visual. The exact coordinates of a micro-relief are deliberately
not published (ethics: do not guide anyone toward a specific site, see the
anti-detecting disclaimer above).

## Documentation

- **User README**: this file
- **Build & deployment**: [BUILD.md](BUILD.md), bundle architecture, per-OS build scripts, updating without rebuild, troubleshooting (including Linux- and macOS-specific cases)
- **Built-in help**: `python lidar2map.py --help` (LiDAR), `--raster --help` (raster), `--vector --help` (vector), `--osm --help`, `--merge --help`

## License

Code distributed under the **GNU General Public License v3.0**, see [LICENSE](LICENSE).

You are free to use, modify and redistribute this software under the terms of the GPL v3. In particular: if you redistribute a modified version, you must provide the modified source code under the same license.

## Author

Designed and architected by **Nicolas Martin** ([@nico579](https://github.com/nico579)). Code developed with the assistance of Claude (Anthropic) as a development tool.

## Acknowledgements

Data used:
- **IGN** (French National Institute of Geographic and Forest Information), LiDAR HD, BD ORTHO (including the historical 1950-1995 versions), BD TOPO, under the Etalab 2.0 license
- **AHN** (Actueel Hoogtebestand Nederland), AHN4/5 0.5m (Netherlands), CC BY 4.0
- **swisstopo** (Swiss Federal Office of Topography), swissALTI3D 0.5m (Switzerland), free open data © swisstopo
- **Kartverket**, Nasjonal Høydemodell 1m (Norway), CC BY 4.0
- **Geobasis NRW · LDBV Bayern · LGLN Niedersachsen · TLBG Thüringen**, DGM 1m (1-2m Thuringia) (Germany, 4 Länder), Datenlizenz Deutschland Namensnennung 2.0
- **Land Tirol** (tiris), DGM 0.5m (Austria, Tyrol), CC BY 4.0
- **Environment Agency** (England) & **DataMapWales / Natural Resources Wales**, LIDAR Composite DTM 1m (UK), Open Government Licence v3
- **Scottish Government / JNCC** (Scottish Remote Sensing Portal), Scottish Public Sector LiDAR DTM 0.5m (Scotland), Open Government Licence v3
- **ACT** (Administration du Cadastre et de la Topographie), BD-L-Lidar 2024 DTM 0.5m (Luxembourg), CC0
- **USGS**, 3DEP / The National Map 1m (USA), public domain
- **GSI** (Geospatial Information Authority of Japan), DEM5A elevation tiles 5m (Japan), GSI content terms
- **Digitaal Vlaanderen**, DHMV II DTM/SVF/Hillshade (Belgium Flanders), Open Data Licentie Vlaanderen
- **Maanmittauslaitos**, Elevation Model 2m (Finland), CC BY 4.0
- **Klimadatastyrelsen / Datafordeler**, DHM DTM 0.4m (Denmark), CC BY
- **Geological Survey Ireland**, LiDAR DTM 1m (Ireland), CC BY 4.0
- **Natural Resources Canada**, HRDEM Mosaic 1m (Canada), Open Government Licence
- **ČÚZK** (Czech Office for Surveying, Mapping and Cadastre), DMR 5G 1m (Czechia), Open Data
- **IGN España / CNIG**, MDT 5m (Spain), CC BY 4.0
- **ICGC** (Institut Cartogràfic i Geològic de Catalunya), MET LiDAR 50cm (Catalonia), CC BY 4.0
- **GUGiK** (Polish Head Office of Geodesy and Cartography), NMT 1m LiDAR ISOK (Poland), open data
- **LINZ** (Land Information New Zealand), 1m DEM (New Zealand), CC BY 4.0
- **QSpatial** (State of Queensland) & **Spatial Services NSW**, 0.5m / 5m DEM (Australia), CC BY 4.0
- **Geoscience Australia**, DEM of Australia derived from LiDAR 5m (Australia, national), CC BY 4.0
- **OpenStreetMap**, vector data under the ODbL license, distributed by Geofabrik
- **Apache JMapsforge / mapsforge-map-writer**, offline vector rendering engine

Bundled tools: GDAL, osmosis, py7zr, pyproj, numpy, scipy, Pillow, ijson, pywebview.