Data
The open data, and what each source is used for
The world is generated from openly licensed data only. Four sources feed the generator today:
| Source | Used for |
|---|---|
| Geoscience Australia, 5 m elevation model from LiDAR | Ground height |
| OpenStreetMap | Road centrelines and classes, one-way streets, lanes, roundabouts, parks, schools, bus stops, paths, water |
| Main Roads Western Australia open road layers | Road hierarchy, and telling a right of way from a car-park aisle (in the suburb prototype) |
| Microsoft building footprints | Building outlines and heights |
Things learned while joining them:
- OpenStreetMap is not enough for buildings in ordinary suburb. One established suburb has 4 buildings in OpenStreetMap where the Microsoft set has 866. Where OpenStreetMap is complete, its count and area match Microsoft's to within 5 %. So footprints come from Microsoft everywhere.
- The two road sources are joined by nearest segment. OpenStreetMap gives the geometry; the Main Roads hierarchy overrides the class where more than half the sample points match. 97 to 100 % of road ways found a match.
- Lots are inferred by subdividing street frontage, guided by the footprints: 983 lots against 852 dwellings in the measured square kilometre. About one in six holds no house, mostly corner fragments, so fences are over-counted by 10 to 15 %.
- Heights give storeys: a footprint taller than 5.8 m is two storeys; a footprint under 60 m² is an outbuilding.
Licences and the attribution each source asks for are on the credits page.
There is no open tree-canopy data and no open lot boundaries, so tree densities are stated assumptions and lots are inferred. No source has been checked against the ground.