3.2.1. Concept

Emissions from individual vehicles can be estimated with vehicle trajectory data collected from GPS devices by the method described in Section 3.1. The problem is that the total amount of emissions cannot be calculated because it is not possible to collect the driving trajectory data from all vehicles in the target area. One way to consider is to make the extracted vehicle trajectories from the results of the microscopic tra ffic simulation network model, which can replace the actual vehicle trajectories. However, it has several disadvantages. First, it is not easy to calibrate a tra ffic simulation network model to make it similar to the actual tra ffic network. Second, if the target area is changed, the method can be applied only after establishing a new simulation network for that area. Furthermore, even if this method enables all the vehicle trajectories of the entire network to be acquired, calculating emissions from all vehicles in a large network using the micro-level emission estimation method would require a lot of time.

Another alternative is to find representative vehicle trajectories on the target road section at the analysis time interval for each vehicle type (passenger cars, trucks, buses), calculate emissions from those vehicles' trajectory data, and use the calculation results to estimate emissions from all vehicles in the road section at the analysis time interval. Because the tra ffic volume for each vehicle type in the analysis time interval of the target road can be obtained by using ITS equipment or the tra ffic volume estimation model, the total emissions can be calculated by multiplying the tra ffic volume by the emissions from the representative vehicle trajectories. This method has the advantage of higher efficiency in calculating regional vehicle emissions because the calculated emissions based on the representative vehicle trajectories from each road section can be used as emission factors (g/veh) for each road section. On that basis, in this study, this method has been applied to extract the trajectory of a vehicle, which has representative driving characteristics, from the vehicle trajectory data of some vehicles that passed the corresponding road section at the corresponding time interval.

Several studies aimed at developing representative vehicle trajectories have been established. For example, a representative link driving cycle by vehicle type, road type, and LOS has been developed for MOVES. However, it is not appropriate to apply their driving cycles to estimate regional vehicle emissions. The first reason is that the concept of the measurement link of the MOVES driving cycle is di fferent from the link in the transportation network of the node link system. The former is closer to the travel route concept, while the latter refers to a road section whose length can vary and is shorter than a travel route. The second reason is that the concept of driving cycle is di fferent between the MOVES driving cycles and the extracted representative link vehicle trajectories in this study. The MOVES

driving cycles are developed to represent vehicle driving characteristics of various road types. On the other hand, this study extracts the representative vehicle trajectories to reflect the various driving situations that can appear in each link based on the driving data collected from each link. In other words, the driving characteristics of the same vehicle type that runs on the same road at the same time interval should have many similarities, as well as certain differences. For example, in the case of an arterial road, there will be a difference in vehicle trajectories between a vehicle experiencing traffic delay due to signals and one that is not. Thus, this study intends to classify vehicle trajectories into several similar groups and use the center of each group as a representative vehicle trajectory.

Figure 1 is a diagram showing the procedure for estimating the micro-level link emission factors required to calculate the emissions at the network level by using the collected vehicle trajectory data. The following section explains each process in detail.

**Figure 1.** Process of Developing Link Emission Factors.

3.2.2. Calculate OpMode Distribution of Vehicle Trajectory

First, the vehicle trajectories of the same vehicle type on the same road section at the same time interval are collected. The vehicle trajectory data includes link ID, vehicle ID, recording time, and speed at each time point (in sec). Second, the VSP per second is calculated by the method described in Section 3.1.2 for each vehicle trajectory, and OpModes per second are classified for each vehicle trajectory. Then, the frequency for each of the 23 OpModes is calculated, and the ratios of each OpMode are calculated and stored. Each vehicle trajectory has 23 new variables, which are the ratios of each OpMode.
