A GeoAI framework for remote-sensing-driven traffic activity and Emission Intensity in a data-scarce megacity

Key Points

• This research aims to develop a GeoAI framework to assess traffic activity and emissions in Kinshasa using remote sensing data.
• Utilized YOLOv8 for vehicle detection from satellite imagery, establishing vehicle counts.
• Applied a gravity model incorporating land-use data and bus traffic data to distribute vehicle volumes across a road network.
• Conducted emission calculations based on estimated speeds and traffic volumes for NOx and PM2.5 on a road-segment level.
• Successfully generated hourly emission maps for NOx and PM2.5 across different road segments.
• Achieved accurate vehicle counts with YOLOv8 resulting in reliable bus-to-private vehicle ratios.
• Validated the gravity model effectively distributed vehicle volumes based on land-use classifications.