Calibration and validation of a micro-simulation model in network analysis

This paper documents the calibration and validation efforts carried out as part of network analysis of a sub-area in the City of Niagara Falls using the Paramics micro-simulation model. The calibration effort involved comparing the model results to the field data that included not only link traffic volumes and turning movement counts at intersections but also measures of effectiveness such as average travel times and approach queues. Paramics uses a dynamic assignment procedure in which movements of vehicles through the network are governed by origin-destination matrices on the basis of various assignment techniques. For that reason the modeling exercise involved estimation of suitable origin-destination matrices which could replicate the observed traffic volumes and turning movement counts at selected intersections to acceptable levels. Target benchmarks were chosen and used as the basis of comparison between modeled and observed volumes using a modified Chi-Squared statistic test. Further model validation was conducted by comparing modeled and observed measures of effectiveness. It was found that target benchmarks that demonstrated an acceptable match between modeled and observed results were achieved with moderate calibration efforts. However, greater efforts are required to achieve marginal improvements in the accuracy of the model outputs and the ability of the model to predict the measures of effectiveness largely depended on the closeness of the match between observed and modeled traffic volumes. The study demonstrates that microsimulation can be applied successfully to network analysis but notes that detailed data is required to conduct the calibration and validation of the model successfully.