The main objectives of this research are; to develop on-road emission factor estimates for carbon monoxide (CO) and hydrocarbon (HC) emissions; to collect traffic and vehicle parameters that might be important in explaining variability in vehicle emissions; to develop an empirical traffic-based model that can predict vehicle emissions based upon observable traffic and vehicle parameters. Remote sensing technology were employed to collect exhaust emissions data. Traffic parameters were collected using an area-wide traffic detector, MOBILIZER. During the measurements, license plates were also recorded to obtain information on vehicle parameters. Data were collected at two sites, having different road grades and site geometries, over 10 days of field work at the Research Triangle area of North Carolina. A total of 11,830 triggered measurement attempts were recorded. After post-processing, 7,056 emissions were kept in the data base as valid measurements. After combining with the traffic and license vehicle parameters, a data base has been developed. Exploratory analysis has been conducted to find variables that are important to explain the variability of the emission estimates. Statistical methods were used to compare the mean of the emissions estimates for different sub-populations. For example, multi-comparison analysis has been conducted to compare the mean emissions estimates from vehicles having different model years. This analysis showed that the mean emissions from older vehicles were statistically different than the mean emissions estimates from the recent model year vehicles.One of the contributions of the research was developing an empirical traffic-based emission estimation model. For this purpose, data collected during the study were used to develop a novel model which combines the Hierarchical Tree-Based Regression method and Ordinary Least Squares regression. The key findings from this research include: (1) the measured mean CO emission estimate for Research Triangle park area of North Carolina is estimated as 340 grams/gallon, whereas the mean HC emissions estimate is found to be as 47 grams/gallon (2) inter-vehicle variability in vehicle emissions can be as high as two orders-of-magnitude; (3) intra-vehicle variability is lower compared to the inter-vehicle variability; (4) some vehicle variables such as vehicle model year and vehicle type are important factors in explaining the inter-vehicle variability in emissions estimates; (5) emission estimation model developed in this research can be applied to estimate the emissions from on-road vehicles.
