Finding Byzantine junctions with Steiner trees

The Byzantine road system in Cappadocia (Turkey) is known in reasonable detail. Reconstructions of the road system by Hild (1977) show the most probable location of the routes that were built and maintained by the imperial administration. Unlike the earlier Roman routes, these were not only built to establish fast connections between major cities; rather, they were intended to connect all administrative and religious centres in the area. A conspicuous element of this connecting network is the fact that road junctions are found in the countryside, without any corresponding traces of settlement. Least cost path models trying to connect the centres will inevitably fail to identify these junctions, since LCP modelling will only create routes between single sources and targets, and not find the optimal network solution to connect three or more destinations. Steiner trees are an appropriate mathematical solution to deal with this problem; however, applications in GIS are lacking up to now. In this paper we would like to present the preliminary results of a case study showing that it is technically possible to use Steiner trees in conjunction with cost surfaces. The modelling results seem promising, as the resulting connections are closer to the documented ones than those resulting from least cost path modelling. However, there are substantial unresolved issues regarding the implementation of this method in GIS.