Improved road traffic emission inventories by adding mean speed distributions

Abstract

Does consideration of average speed distributions on roads—as compared to single mean speed—lead to different results in emission modelling of large road networks? To address this question, a post-processing method is developed to predict mean speed distributions using available traffic data from a dynamic macroscopic traffic model (Indy) that was run for an actual test network (Amsterdam). Two emission models are compared: a continuous (COPERT IV) and a discrete model (VERSIT+ macro). Computations show that total network emissions of CO, HC, NO x , PM 10 and CO 2 are generally (but not always) increased after application of the mean speed distribution method up to +9%, and even up to +24% at sub-network level (urban, rural, motorway). Conventional computation methods thus appear to produce biased results (underestimation). The magnitude and direction of the effect is a function of emission model (type), shape of the composite emission factor curve and change in the joint distribution of (sub)-network VKT (vehicle kilometres travelled) and speed. Differences between the two emission models in predicted total network emissions are generally larger, which indicates that other issues (e.g., emission model validation, model choice) are more relevant.