Evaluation of satellite-derived building height extraction by CFD simulations: A case study of neighborhood-scale ventilation in Hong Kong

Computational fluid dynamics (CFD) techniques, such as large-eddy simulation (LES), are widely used in urban ventilation studies. Unfortunately, realistic building data required in CFD studies are not always available. In this study, LES of urban ventilation is used to assess building height extraction from different satellite images. As a case study, three sets of digital elevation data extracted from satellite images in an urban area of Mong Kok, Hong Kong, are assessed. Simulations of velocity ratio from the LES model are first validated by wind tunnel measurements, yielding a recommended local roughness length of 0.02 m. Wind characteristics in urban elevations extracted from two single satellite images and the fused result are then compared with those from realistic data in the same area, using identical LES settings. It is found that building height data retrieved from the WorldView-2 optical (stereo) images are of poor quality, as they underestimate higher buildings, which are deemed to be more important for pedestrian-level velocity ratios. The TerraSAR-X synthetic aperture radar (SAR) image and the fused results of SAR and stereo images are sufficient for CFD simulations of urban ventilation. This fusion provides a slight improvement over SAR images alone in reproducing urban geometric parameters, and it is superior in simulating more accurate wind profiles in the urban canopy.