**6. Discussion**

The applicability of the developed methodology was examined by using the navigation data collected according to each of the three (morning peak hours, non-peak hours, afternoon peak hours) analysis time intervals on one highway section (1 km) and one arterial road section (600 m). The results of the analysis confirmed that the error rate showed a difference in the range of 1–4% for the highway section and 1–6% for the arterial road section when the link emissions were calculated using the proposed method. Moreover, the results indicated that the estimated micro-level link emission factors reflect the driving characteristics of the link according to the traffic conditions for each time period and the driving characteristics for each road characteristic (uninterrupted flow and interrupted flow). Additionally, the analysis results through the same method while accumulating the vehicle trajectory data showed that clustering analysis produced similar cluster centers after 4-day data accumulation and link emission factors were converged to certain values representing the vehicle travel characteristics on the corresponding road at the corresponding time period by adding the number of data collection days. The number of days to reach convergence on the target road in this study was about four, while it is

expected to increase in road sections having fewer collected vehicle trajectories. These results mean that the proposed method can offset the limitations of vehicle trajectory data with a small number of samples. Furthermore, this approach enables the analysis time period to be shorter and thus the micro-level emission calculation of the entire network to be analyzed in more detail.

In this study, because navigation data was used, the results were limited to the case of passenger cars. Therefore, it is necessary to acquire the vehicle trajectories of a commercial vehicle, such as a truck or bus, from DTG data to check whether the applicability is the same. In addition, considering that the link length of the traffic network varies, it is also necessary to vary the collected link length to check whether there are any points to be supplemented in the analysis method. Moreover, it is also required to confirm the basis for judgment as to how many days for data accumulation and analysis would enable representative link micro-level emission factors to be developed.
