Rural road segmentation research

Organisation: RAC Foundation

Date of Publication: February 2023

Uploaded to Knowledge Centre: 8 March 2023

For many years the Department for Transport have applied a binary rurality categorisation to Britain’s roads, which assigns each police recorded injury collision to either an ‘urban’ or a ‘rural’ road. This study examines the diverse collection of roads presently categorised as ‘rural’, with the goal of proposing a new typological approach which describes rural highways based on road user experience.

The project set out to:

  • Assemble data for the highway characteristics of ‘rural roads’ as presently defined.
  • Examine how those characteristics could generate more informative definitions.
  • Arrange these definitions into a formal taxonomy with reproducible outputs.
  • Evaluate the taxonomy and seek stakeholder feedback.

A random sample of ‘rural road’ routes was selected, stratified by road class, with total sample length by relative traffic volume according to national statistics. The sample included 434 routes covering a total 1,563km of Britain’s ‘rural road’ network.

The project began by using machine learning to cluster all sampled routes using input data which covered the physical characteristics of the carriageway, the usage of the road, and the local environment. The output from this model grouped the sample routes into eleven clusters, further arranged into seven superclusters.

The machine learning output was then subjected to qualitative review to verify the allocation of routes to clusters, based on human assessment of their attributes and physical characteristics. Following this review, the taxonomy was refined to ten clusters, which were grouped into four superclusters.

A small number of routes from across the whole sample were held back from the qualitative review for use as a blind sample to test the machine learning outputs further. The results of this test were strong, showing high correlation between the model and qualitative analysis assignments.

The four superclusters defined in the final version of the taxonomy were:

  • Principal Roads, split into two clusters Regional Distributors and Main Roads
  • Country Roads, split into three clusters Primary Routes, Secondary Routes, and Tertiary Routes
  • Neighbourhood Roads, split into two clusters Residential Distributors and Village High Streets
  • Winding Roads, split into three clusters Country Lanes, Hill Passes and Remote Roads.

Click the following link to access the report: