Meso-level Dynamic Estimation And Assessment Of Vehicle Emissions For Road Networks

With the socio-economic and urban development, the motor vehicle ownership has been growing rapidly, resulting in the associated traffic congestion problem being increasingly notable, which is further causing such serious traffic environment pollutions that need to be controlled urgently. As the traffic environment pollutions are getting worse, the research on emission control strategies has been evolved gradually from reducing individual vehicle emissions by improving engine or automobile design, to controlling the total vehicle emissions of road networks by improving vehicle driving conditions. The key to the successful implementation of such more dynamic emission control strategies is the capability to accurately and quantitatively evaluate vehicle emissions of road networks in various traffic conditions. At present, the estimation and assessment of vehicle emissions of road networks generally calculates total emissions of networks from a macroscopic perspective, which could not reflect the temporal and spatial variations of emissions on different road segments, thus is unable to provide guidance and assistance to the identification of high pollution road segments and the development of effective control strategies in a timely manner. In this context, the research in this dissertation strives to propose an approach for the dynamic estimation and assessment of vehicle emissions of road networks from a mesoscopic perspective by utilizing the measured traffic flow characteristics from the real road networks.The estimation and assessment of vehicle emissions of road networks involves four types of data:traffic volume, speed, vehicle type composition, and vehicle emissions. As the present technologies in the automatic collection of traffic data can provide traffic volume and speed in a large scale road network with high temporal and spatial resolutions, the research in this dissertation focuses on two areas, predicting vehicle type composition of the road segment, and predicting emissions from on-road vehicles.First, the vehicle type composition data on typical road segments are collected by using the license plate recognition technique. Then, the major feature indicators of road segments that influence the vehicle type composition are identified by utilizing the decision tree method. Further, in light of the characteristics that the fraction of each vehicle type is between zero and one with the sum of all fractions being equal to one, an approach of multinomial logistic regression is adopted to develop the prediction model for the vehicle type composition, which expands the vehicle type composition for limited road segments to all road segments in the network.Second, the vehicle activity and emission data are collected by the portable emission measurement system (PEMS). Then, vehicle driving conditions are divided by introducing the variable of the Average Speed Increment (ASI). Subsequently, two mesoscopic vehicle emission models are developed by using the average-speed-based method and the VSP-Bin-distribution-based method. Further, a comparative analysis is conducted on the two models from perspectives of model accuracy, data requirements, and applicabilities, after which the best model is selected.Finally, the dynamic estimation and assessment approaches are developed for vehicle emissions from the perspectives of road segments, road, as well as road networks, by combing the prediction models for the vehicle type composition, the estimation model of vehicle emissions, and the real measured traffic flow data. Then, by using Beijing as the case study city, the bottom-up aggregation method is adopted to analyze and evaluate the temporal and spatial distribution of vehicle emissions as well as the current pollution status for the road network within the 5th Ring Road Expressway.