DVRPC’s new regional travel forecasting model has several new features that improve transit modeling and analysis. These include an improved representation of the transit network using the General Transit Feed Specification (GTFS) data provided by transit providers. The transit network is accurately modeled with regard to route alignment, stop locations, service schedules, and fare information. The model has been validated to ensure that it could accurately represent travel patterns at the regional level. However, the model validation process rarely includes tests to see how well regional models respond to changing conditions. In particular, how well does the model do at estimating a change to key inputs, such as transit fares?

To specifically address this issue, a study was designed to evaluate the regional model’s ability in predicting the effect that transit fare policies at SEPTA, one of the major transit operators in the DVRPC region, have on ridership. In this study, a set of fare change scenarios, including both hypothetical and “real-world” situations, were modeled. For the hypothetical scenarios, fares in the model were changed in an arbitrary way (across-the-board increase for instance), and then the resulting change in ridership as estimated by the model was compared to the expected change according to the literature. The real-world scenarios involved two actual fare changes as implemented by SEPTA. In both cases, the same exact fare change was entered into the model, and other key factors (such as gas price) changed from the base year (to which the model was calibrated) was also modeled. Then the resulting ridership change as estimated by the model was compared to the actual observed change in ridership as reported by SEPTA. This presentation is intended to share the experience and findings of this study - how backcasting and sensitivity analysis could help the validation and improvement of the regional travel forecasting model.