Network data sources

Every accessibility study starts with a decision about network data. The choice constrains which modes can be modelled, which areas can be covered, and how much cleaning work follows.

Comparison

Source	Coverage	Modes	Licence	Main limitation
OSM	Global	Car, bicycle, pedestrian, rail	ODbL	Variable completeness and tagging quality
NWB	Netherlands	Car, with cycle and foot paths in a separate file	Open	National only, so cross border routes are not represented
GTFS	Per transit operator or region	Public transport	Varies	Timetables only, no infrastructure geometry
Commercial (TomTom, HERE)	Global	Car	Proprietary	Cost, but it includes speed profiles by time of day

Sources are routinely combined: OSM geometry for cycling and walking, a commercial source for car speed profiles reflecting congestion, and GTFS for transit. The Dutch national network model does exactly this. The worldwide network model takes the opposite approach and builds a car network for the entire planet from OSM alone, per continent, combined with the GHSL population grid and settlement layer.

What to check before committing to a source

1. Modes. Does it contain cycle and foot paths, or only roads a car can use? A car only file makes pedestrian accessibility impossible, not merely approximate.
2. Extent and edges. Does it cover the area around your study area? Many fastest routes within the Netherlands run through Belgium or Germany, and a national file will route them the long way. See Zoning and spatial units on edge effects.
3. Directionality and restrictions. One way flags, turn restrictions, and access tags for private roads, bus lanes and pedestrian zones.
4. Connectivity. Whether links that cross actually share a node, see Building a routable network. Bridges and tunnels must not connect.
5. Attributes for speed. Road classification, speed limits and surface, the inputs to Travel speeds per mode.
6. Update frequency and versioning. A reproducible study needs a dated snapshot, not "current".

Related

• Building a routable network, turning a data file into a graph
• Travel speeds per mode, assigning impedance to links
• Euclidean versus network distance, what you lose without a network