| Summary: | Motor vehicle emissions are a major air quality concern in Chinese cities. However, traditional population-based emission inventory methods fail to capture the spatial and temporal variations in emissions for effective policy design. This study proposes a high-resolution approach for traffic flow prediction and vehicle emission inventory estimation, using Jinan City, China, as a case study. We leverage multi-source geospatial data and employ a two-fold random forest model to predict hourly traffic flow at a road-segment level. Speed-aligned emission factors were then combined with these data to calculate hourly and road-level vehicle emission estimates. Compared to traditional methods, our approach offers substantial improvements: (1) improved spatiotemporal resolution; (2) enhanced accuracy of traffic flow prediction; and (3) support for more effective vehicle emission control strategies. Results show that heavy-duty vehicles, particularly freight trucks operating on inter-regional corridors through Jinan, contribute 78% more to NO<sub>X</sub> emissions than local light-duty vehicles. These transient emissions are typically overlooked in static inventories but constitute a significant source of urban pollution. This study offers valuable insights for combining geospatial data and machine learning to improve the accuracy and resolution of vehicle emission inventories, supporting urban air quality policy and planning.
|
|---|
