Proper orthogonal decomposition of large-eddy simulation data over real urban morphology

Proper orthogonal decomposition (POD) is implemented to the atmospheric surface layer (ASL) over real urban morphology based on the data of large-eddy simulation (LES) with explicitly resolved buildings. Turbulence structures with different scales in the inertial (ISLs) and roughness (RSLs) sublayers are unveiled by decomposing the fluctuating velocity components into a range of spatial eigenmodes psi m(x) and time coefficients am(t). POD spatial modes, which are ranked according to their turbulence kinetic energy (TKE), are visualized to contrast their sizes. These dominant structures illustrate the footprints of most energetic eddies at the ground level which could modify the dynamics locally. Moreover, their temporal behaviors are examined by energy spectrum and Lissajous curve, elucidating the existence of large-scale traveling structures. It is also shown that around 80 and 100 modes (out of 1000) are able to retrace 90% of the original TKE and momentum flux, respectively. The aforementioned remarkable data reduction demonstrates the capability of POD for flow characteristic extraction and city-scale, reduced-order model (ROM) development. Besides, dominant modes transport more efficiently especially in ISLs. With proper urban planning, such as building height and orientation, these energy-carrying eddies would sweep into street canyons, improving pedestrian-level air quality as well as city ventilation. (200 words)

Automated summary

The study implemented Proper orthogonal decomposition (POD) to analyze turbulence structures in the atmospheric surface layer (ASL) over real urban morphology, revealing potential impacts on urban wind risks and air quality. The results demonstrate the high capability of POD in reducing data dimensions and extracting flow characteristics, providing new insights for urban planning and air quality improvement.