It is often discussed among practitioners how many zones to apply in a traffic model. The more zones the better details in the modelling. However, more zones also require more data and more calculation time - the latter approximately increasing with the square of the number of zones. Even though computers get faster, the latter is still an issue, e.g. if more advanced stochastic user equilibrium assignment models are used. The OTM-model ofthe Copenhagen Region has been developed over a series of years. The mostrecent version has a quite detailed level with regard to the zonal structure, as it has 835 zones. The model covers the Copenhagen Region and its hinterland. The Oeresund model is a freight transport model covering the Southern part of Sweden an East Denmark. This model has 1441 zones. Finally,the European Transtools model is used, where the TEN-CONNECT version has 832 zones. The three models form a good test bed with regards to zonal structure for different geographical scales of models. The models also have arich database of traffic counts. The paper analyses the influence of the zonal structure on the results running tests by these three models. The benchmark uses the most detailed versions of the model. Based on these a method was developed to merge zones and matrices, and create new zone centroids and zone connectors from the zones to the network. Hereby, it is possible to analyse the difference of results by aggregating the zonal structurecompared to the benchmark version, as well as the changes of fits with traffic counts. Also calculation time is analyzed. Finally, it is analyzed how the models perform for project appraisal (selected projects) using different zonal structures. The study presents the results of the analyses, and try to provide some guidelines with regard to suitable aggregation of zonal structures for the three types of models; metropolitan, national/regional and international. For the covering abstract see ITRD E145999
Samenvatting