Traffic simulation calibration is often a complicated process which involves adjusting network parameters and other input data so that the simulation results provide a better match to observed conditions. This should be a common practice amongst planning organizations or consulting firms. However, if an agency is less familiar with mesoscopic models, and most of their experience has been working with macroscopic models, their traditional static assignment-based calibration approaches might not be comprehensive for capturing time-dependent traffic congestion. Additionally, mesoscopic calibration may require a combination of model adjustments which are obvious to neither new professionals nor experienced planners. What happens if we adjust the traffic flow model parameters at bottlenecks or non-active bottlenecks? Maybe we need to re-evaluate our link-specific capacity attributes? Could we make joint modifications to the OD demand matrix and bottleneck capacity? What happens if we only change path flow distribution or departure time patterns?

The purpose of this paper is to provide some helpful guidance/details to practitioners about how to calibrate mesoscopic traffic simulation models, with real-world examples selected from multiple regions. We’ll examine the effects of changing different modeling parameters in mesoscopic simulations, which parameters are most sensitive to adjustments, and which parameters we can reasonably justify changing, all while considering how the traffic assignment and network loading components function in the simulation. Beyond modeling parameters, this paper will also discuss agent-based joint Origin-Destination Matrix and Path Flow Estimation (ODPE) techniques for application in the model calibration process. Specific examples will be shown with both large regional networks and subarea networks for more detailed analyses.