**1. Introduction**

Emissions from on-road mobile sources depend on the driving characteristics of the vehicles. The emission calculation methods are largely divided into macro-level approaches that consider the average speed of vehicles traveling along the roads as a driving characteristic and micro-level methods that reflect the change in instantaneous speed of individual vehicles as a driving characteristic [1]. Since the data on average speed, distance traveled (or link length), and traffic volume on the road section, which are necessary input data for calculating macroscopic emissions, is basically built and managed in the traffic network data, it is relatively easy to calculate the total emissions of the traffic network. For this reason, the macro-level emission estimation method has been generally applied and utilized for calculating the emissions from on-road mobile sources in different regions and countries [2–4]. However, in the macro-level emission estimation method using the average speed as the deterministic variable, the amount of vehicle emissions may be under- or overestimated because this method cannot capture the instantaneous changes in the vehicle driving speeds, such as stop-and-go traffic situations [5,6]. For this reason, there have been many claims that the micro-level emissions method should be applied in estimating regional and national emissions.

A number of previous studies have mounted GPS equipment onto only experimental vehicles to collect the vehicle trajectory data and estimate the vehicle emissions [7–9]. However, with the advancement of electronic communication technology, vehicle trajectories are being continuously collected nationwide by using vehicle navigation devices, digital tachographs (DTGs), and mobile devices, which have become available for estimating vehicle emissions [10–12]. The collection of vehicle trajectory data has become easier, and the range of spatio-temporal data collected has been expanded. As a result, it is possible to estimate micro-level emissions from the collected vehicle trajectory data.

However, there are some practical di fficulties associated with adopting the micro-level approach at the regional and national levels. First, developing micro-level emission factors requires a large amount of time and cost. For vehicles with various fuel types, fleet sizes, and model years, it is necessary to conduct multiple driving tests, measure emissions under various operating conditions, and derive emission factors suitable for the micro-level emission estimation method. A fast solution is to use micro-level emission factor databases from other countries. In fact, the U.S. Environmental Protection Agency's (U.S. EPA) Motor Vehicle Emission Simulator (MOVES) [13] can estimate micro-level vehicle emissions, and it has been applied in various studies from many countries [14–18] even though the approach should be taken with caution because the classification method for vehicle types and emission standards vary by country. The second problem is related to the acquisition of vehicle trajectory data, which is required to calculate micro-level vehicle emissions. In the case of collecting the link average speed of tra ffic flow, the average speed data can be easily collected from tra ffic information systems, such as loop detectors. However, estimating vehicle emissions with micro-level emission models is limited because it requires second-by-second vehicle trajectory data. It would be ideal to collect the trajectories of all vehicles driving along the roads to estimate the emissions of local or national on-road mobile sources on a micro-level basis, but that is highly impractical. Therefore, it would be useful to apply vehicle emissions, which are estimated at a micro level, at the regional and national levels with an easier and faster method.

This study proposes a representative vehicle trajectory extraction method for estimating micro-level vehicle emissions with a limited amount of vehicle trajectory data, such as that from DTGs or mobile devices. In the method, MOVES is used for analyzing vehicle emissions at a micro level, and vehicle trajectory data is divided into several groups through a k-means clustering method, in which the ratios of each operating mode (OpMode) in MOVES are used as cluster variables for clustering similar vehicle trajectories.

The rest of this paper is organized as follows: Section 2 describes the uniqueness of the representative vehicle trajectory extraction methodology used in this study, and Section 3 explains the proposed network-level micro-level emission estimation procedure, representative vehicle trajectory extraction method, and micro-level emission factor derived from this study. Section 4 presents the results of applying the proposed method to navigation data collected in Bucheon, Gyeonggi-do in Republic of Korea. Section 5 presents the e ffects of using the accumulated vehicle trajectory data on the method. In Section 6, the implications learned from the analysis results and the limitations of this study are discussed. The conclusions for this study are mentioned in Section 7.
