Calibration and validation of a dynamic traffic assignment (DTA) model present new and important challenges to the planning community. Given that the perceived value of DTA for planning is in DTA’s promise to better capture traffic flow dynamics, and thus to produce a more accurate picture of the impacts of projects, it is imperative that a DTA model be calibrated and validated to observed dynamic, or time-varying, data – e.g., five-, 10-, or 15-minute volumes, speeds, queue lengths, or other measures of performance. These data, while not commonly found in a metropolitan planning organization’s data catalog, are potentially on offer from traffic data vendors such as INRIX. In this paper, we explore the use of INRIX’ real-time speed data in the validation of a wide-area microscopic simulation and DTA model of Central Phoenix.

The principle motivation behind the calibration and validation of a DTA is to, at a minimum, be able to replicate observed bottlenecks when they occur, where they occur, and with the intensity and duration that are observed in the field. To achieve this validated result – that is, to produce a model that reflects the desired congestion patterns spatially and temporally – hinges largely on a reasonably accurate representation of the traffic demand (i.e., the spatial and temporal distribution of trips). Thus, to validate the model is to make adjustments to the trip table that bring the DTA model in better agreement with the observed validation data while simultaneously improving, or not significantly diminishing, the model’s goodness-of-fit with the calibration data. In the model of Central Phoenix that we present, the model is calibrated to traffic count data at a 15-minute resolution, and we present a methodology for using INRIX speeds, also reported at a 15-minute resolution, to steer the model’s validation.

We present validation results depicting the model’s ability to match key bottlenecks in and around Central Phoenix. We also describe the data challenges one faces in fusing INRIX data with simulation networks, interpreting INRIX’s speed data, and comparing INRIX’s speed data with DTA results. Finally, we weight the potential and challenges for developing an automated validation methodology in future applications.