An Analysis of Zone Centroid Location Impact on Traffic Assignment Results; Case Study of Mashhad, Iran

Transportation planning studies are generally based on dividing the study area into a set of traffic analysis zones (TAZ) based on a set of rules. Each zone is then represented by a node named centroid, which is assumed to be the origin and also the destination of all trips of that zone. Considering the sensitive and focal role of the location of centroids on traffic assignment results due to this rather big assumption which may cause large errors in results, very little research is found in the literature. Due to this research gap in this field, the current paper tries to analyze and examine the impact of centroid location on traffic assignment results. The traffic assignment applied is that of Wardrope’s user equilibrium (UE) principle implemented in VISUM transportation planning software package and the quantitative measures are those of hourly flows on different links of the transportation network. Predicted link flows for different scenarios are compared to the observed flows on the same selected links. To discuss and analyze the implementation challenges, the real sized network of the second largest city of Iran, namely the city of Mashhad is selected as the case study. Three different methods for selecting centroids are presented to analyze their impacts on link flows. For each method, two different scenarios of choosing connector nodes are presented: in the first scenario, number of connector nodes is variable for different zones but their locations are fixed for each of the three methods. In the second scenario, having chosen the centroid according to each of the three methods, the closest node to the centroid is selected as the connector node. Results of this research, based on the comparison of predicted and observed flows on 87 links of Mashhad transportation network, show that for the assumptions of this case study, changing centroid location can cause a change in the coefficient of determination of the linear regression line of predicted-observed flow links of 1.7 percent in the first and 20 percent in the second scenario of selecting connector nodes. This noticeable impact in a real case with realistic assumptions can be considered an important confirmation of such research and also the need for further research in this rather unattended field.