Key points are not available for this paper at this time.
Data on road traffic speeds is needed for air and environmental noise pollution modelling, for regulatory control, exposure assessment and assessing health impacts. However, vehicle speed data is often not available and national speed limits, by road classification, may be used as a proxy instead. This may contribute to uncertainties in model predictions. This study presents novel methods, applied to satellite imagery, to calculate vehicle speed road by road for entire cities, demonstrating their use in Barcelona. The approach exploits the fraction of a second gap between eight multi-spectral sensors onboard the WorldView-2 and -3 satellites and was used to estimated speeds for 128,206 vehicles on motorways, trunk, primary, secondary and tertiary roads from 10 satellite images. The average estimated vehicle speeds ranged from 69 km/h on motorways to 31 km/h on tertiary roads. Satellite-derived vehicle speeds showed good agreement with Directions API speeds on a subset of roads (R 2 = 0.71, NMGE = 0.23 and RMSE 13.4 km/h). A Normalised Mean Bias of 0.04 suggests that on average the satellite and Directions API estimates were similar. Using Directions API as the benchmark, satellite-derived estimated speeds yielded values of RMSE of 13.4 km/h, compared with and RMSE of 25.7 km/h using road type national speed limits, for a subset of roads. In this pilot study, therefore, satellite-derived vehicle speeds yield a 48% reduction in error over using the national speed limit by road type. This paper shows that high-resolution satellite imagery has potential to quantify vehicle speed in cities. • A novel method for estimating road traffic speed using high-res satellite imagery. • Methodological developments in data cleaning and quality checking. • Estimated vehicle speeds yielded a 48% lower error than road type speed limits. • Good agreement (R 2 = 0.71) between estimated speeds and measured speeds.
Sheehan et al. (Wed,) studied this question.