The impact of centroid connectors on transit assignment outcomes

In transit modeling, access and egress conditions are often overlooked. The most common modeling technique of these conditions relies on the use of centroid connectors. This definition often uses the geographic position of zone centroids and sets constraints on the maximum number and length of connectors. This definition is subject to spatial aggregation issues and has already been proven to bias car assignment outcomes. The impact on transit assignment outcomes has not yet been demonstrated. The current paper investigates the statistical impact of connectors on transit assignment outcomes in an urban model of Lyon in France. Findings suggest that transit ridership, total passenger-kilometers and transit transfers are dependent on the definition of centroid connectors. Setting arbitrary values for the maximum number and length of connectors statistically affects transit results. The pattern and magnitude of this impact vary, however, between transit modes. The bus and rapid bus systems have been shown to be more sensitive towards the definition of connectors than the subway and the light rail systems. These findings question, to a certain extent, the validity and reliability of transit modeling outcomes.