The characterisation of track substructure is an integral component of track maintenance and
rehabilitation. Traditionally this is done by determining the geotechnical properties of the
discrete track formation layers by excavating test pits and sampling the substructure layers.
The development of geophysical investigation techniques such as ground-penetrating radar
(GPR) allows continuous assessment of the condition of the track substructure. The research
described in this paper uses GPR to develop a track substructure characterisation model and
provides classifications for both the ballast and formation layers. The ballast and formation
were classified into four classes, namely very good, good, moderate and poor. When applying
the model to two sections of track (with generally good and poor quality) 82% and 100% of
classes had only one class difference compared to the traditional characterisation tests for the
formation and ballast layers respectively. The GPR track substructure characterisation model
that was developed therefore compares well with traditional characterisation techniques
and will result in significant cost and time reduction. The characterisation of the track using
GPR provides a continuous classification and enhances the accuracy of the data on which a
maintenance engineer can base decisions.