Machine learning based estimation of urban on-road CO2 concentration in Seoul

As it is predicted that the amount of urban on-road CO₂ emissions will continue to increase, it is essential to manage urban on-road CO₂ concentration for effective urban CO₂ mitigation. However, limited observation of on-road CO₂ concentration prevents the full understanding of the variation of urban on-road CO₂ concentration. Therefore, in this study, a machine learning based model that predicts on-road CO₂ concentration (CO₂^traffic) was developed for Seoul, South Korea. This model predicts hourly CO₂^traffic with high precision (R2 = 0.8 and RMSE = 22.9 ppm) by utilizing CO₂ observation, traffic volume, traffic speed, and wind speed data as main factors. Analyzing the CO₂^traffic data predicted by the model, the high spatiotemporal inhomogeneity of CO₂^traffic over Seoul with 14.3 ppm by time and 345.1 ppm by road was found. The large spatiotemporal variability of CO₂^traffic is related to different road types (major arterial road, minor arterial road, and urban highway) and land-use types (residential, commercial, bare ground, and urban vegetation) where the road belongs. The cause of the increase in CO₂^traffic was different by its road type, and the diurnal variation of CO₂^traffic was different by its land-use type. Our results demonstrate that high spatiotemporal on-road CO₂ monitoring is needed to manage the urban on-road CO₂ concentration showing high inhomogeneity. In addition, it suggests that a model using machine learning techniques can be an alternative for monitoring CO₂ concentrations on all roads without conducting the observation.

This work was supported by Korea Environment Industry &Technology Institute(KEITI) through "Climate Change R&D Project for New ClimateRegime.", funded by Korea Ministry of Environment(MOE) (2022003560006).