Requirements for the Calibration of Traffic Simulation Models

The recent emphasis on utilising advanced technologies to make more efficient use of existing transportation infrastructure, coupled with the continuing advances in desktop computing technologies, has created an environment in which traffic simulation models have the potential to provide a cost-effective, objective, and flexible approach for assessing design and management alternatives. However, the models must be demonstrated to be valid, and they must be adequately calibrated for local conditions. While the processes of verification, validation and calibration are certainly not new, based on the lack of literature on this topic, it appears that the application of these processes to traffic simulation models is not well defined. This paper examines the issues related to primarily the calibration of traffic simulation models. It attempts to answer, or at least clarify, several key questions, including the following: What is model calibration? What measures of performance should be used in calibration? and When is a model adequately calibrated? The discussion of these questions stems from the goal of obtaining model results that are seen as credible, reliable, and useful, particularly by people who are not modellers. The distinct roles and responsibilities of model developers and model users are discussed. Examples are provided to illustrate the model calibration process.