Modelling the Roman road network in the south-west

A recent modelling project has created a probable map of the Roman road network across the counties of Devon and Cornwall, showing how it may have connected with the wider province of Britannia. While the presence of a few Roman roads had previously been identified in the region – most notably after an extensive LiDAR survey was carried out by the Environment Agency between 2019 and 2022 – it is significantly less well mapped than the rest of Britain. This work, then, has the potential greatly to enhance our knowledge of Roman movements in the south-west.

Image: Parcero-Oubiña et al. (2023) Journal of Computer Applications in Archaeology 

César Parcero-Oubiña, from Incipit CSIC in Spain, and Chris Smart and João Fonte, both from the University of Exeter, used LiDAR data obtained by the Environment Agency’s project and combined it with GIS spatial analysis to calculate the most probable positioning of Roman roads in the region. They started with the known locations of major Roman sites – including the main settlements of Exeter (Roman Isca) and North Tawton, as well as permanent military fortifications. Once the ‘nodes’ of the network were identified, the team then had the harder task of determining what criteria the Romans most probably employed for deciding where to build their roads. By using the small amount of roads currently known in the area, they determined that a model of human movement based on terrain slope, proposed by Irmela Herzog in 2013, was the best fit. Some roads, however, did not work with the model, and the team determined that the intricate river network, which would have caused seasonal flooding, may have affected the Romans’ decisions. By adding this into the calculation, the modelled theoretical paths were then in large agreement with the actual sections of Roman roads.

The next step was to extend this model to cover the whole region and then, finally, to incorporate some ‘fuzziness’. This may seem like a bizarre thing to add into a model, but – looking at the terrain – you can see why it is necessary: while a narrow pass between two hilltops might only have one passable route, a path through a wide-open plain could be in many different locations. By adding ‘fuzziness’ to a terrain-based model, this uncertainty can be included in the map.

The team then went back to some of the LiDAR data to look for Roman roads in locations their calculations predicted. Through this they identified 13km of Roman road between the sites of Bury Barton and Rainsbury that the original analysis had missed. Future work using this model could help turn up even more. The results were published in the Journal of Computer Applications in Archaeology: see