Evaluation of simulation model vehicle activity output for traditional and modal emission modeling
Traffic simulation models are designed to replicate the behaviors and interactions between behaviors, to produce a detailed representation of actual real-word networks. The network is modeled in such a way that each vehicle is considered as a separate entity with interaction with its surrounding environment, which includes the freeway geometry and other vehicles. In recent studies, a comparison of simulation software has demonstrated the inadequacies of emissions modeling based on theoretical car-following logic inherent in most traffic modeling software (Zhang, et al., 1998; Wang, et al., 1998; Hallmark, et al., 1999). Most traffic modeling software assumes the same car-following behavior whether the vehicles are experiencing congestion or free-flow speeds (Dijker, et al., 1998). The objective of the research was to evaluate whether vehicle activity output from simulation models represent real-world driving characteristics sufficiently to be used in either average speed or modal emissions modeling. Field data for average midblock velocity were compared to four scenarios simulation model output for two study areas and two study periods (peak and off-peak). VISSIM models average vehicle speeds at midblock poorly when compared to field data. P-values from (p < 0.001 to p = 0.0015) were found when performing t-tests for differences in means. This supported this study's research hypothesis namely that microscopic traffic simulators inaccurately output average speeds at midblock. Analysis between CORSIM and field data showed similar results to what were found using the VISSIM model. The only exception to this was with results from data collected at University Avenue during peak hours. The p-value for the University Avenue Peak CORSIM model (p= .3413) suggests that mean speeds are similar between the simulation output and what was collected in the field. An important caveat should accompany these conclusions, however. Although statistics are provided in support of the hypothesis that simulation models are inaccurate in modeling average vehicle speeds, this may not be substantively significant with respect to emission modeling. Only twice was it observed that the simulation models underestimated average vehicle speeds by more than 5 mph. A difference larger than this may be substantially important when dealing with vehicle emissions output.