Data quality is crucial for the quality of any analyis result. The fanciest application, the most sophisticated analysis routine, nothing can completely compensate weak data quality. For the digital representation of road networks, graphs are used as the geometrical basis. In a GIS attributes are attached to the geometry, either to linear segments (edges) or to nodes or both.
When the quality of a road network dataset is to be assessed, one can distinguish between the attributive and the geometrical quality. After several posts on the attributive quality of road network datasets (see this , this , this or this one), I want to focus on a geometrical quality aspect: topology .
A correct topology in road network datasets is especially important when the graph is used for routing purposes. Only edges which are properly connected to the network can be traversed. It is favorable to assess the quality of the graph in terms of topological accuracy before a dataset is further processed and prepared for routing analysis.
Standard GIS routines – for example in ArcGIS – allow for the validation of topology rules. For example all dangle edges in a line data set can be identified. But the standard routine is not able to distinguish between the “quality” of these dangles. They can either be the end point of a segment in the network or the result of a topological error (most often caused by inaccurate editing).
In order to identify potential topological error I’ve combined a few geoprocessing tools which select those dangles which are close to another edge. The basic workflow is this:
- Identify all dangle points.
- Buffer the dangle points.
- Overlay the buffer and the network.
- Select the buffers which intersect more than one edge.
The result of this routine is not unambiguous, because not all dangle point buffers are necessarily topological errors. In many cases it is necessary to check the geometry individually. What looks like an overshoot might be a short access road etc. But still, the routine leads to helpful approximations. Here is how the result can look like:
Got interested? Here is the ArcGIS model to reproduce the workflow.