Variations in traffic volumes and changes in travel-related characteristics significantly contribute to the level of vehicular emissions. However, in current practice, travel forecasting models rely on steady state hourly averages and are thus incapable of accurately capturing the effects of network traffic variations accurately on emissions. Recent research has focused on the implementation of modal emission models to overcome some of these shortcomings in existing emission rate models. A primary input to modal emission models is the fraction of time spent in different driving patterns. The estimation accuracy, however, is hampered by the application of static travel demand models for predicting driving patterns. There is a real need to evolve alternate methods to accurately predict driving patterns.
This dissertation proposes an approach to predicting driving patterns more accurately by applying different models at the macroscopic and microscopic network levels. The proposed models more accurately estimate the driving pattern by considering a set of Emission Specific Characteristics (ESC) for each network link. Specific ESC considered in this research includes geometric design elements, traffic characteristics, roadside environment characteristics, and driver behavior.
Two different models have been developed in this study to capture the driving patterns at each network level. The first model is designed to capture macro-scale driving patterns (average speed) in a larger network and the second model is designed to capture micro-scale driving patterns. The two models have been developed using structural equations. They have been calibrated, evaluated, and validated using a microscopic traffic simulation model. Analysis of the models reveals that geometric design elements exert greater influence on driving patterns than traffic characteristics, roadway environment characteristics, and driver behavior in the estimation of emissions. This research has concluded that, for congested traffic conditions, the proposed models capture driving patterns more accurately than current practice and, consequently, these models estimate the range of emissions more accurately. Models that estimate time-dependent emissions in the presence of traffic sensor data were also successfully estimated.
