Validating spatial reproduction of large-eddy simulations with PIV datasets for turbulence statistics at pedestrian level of urban canopy

Abstract

For large-eddy simulations (LES) at pedestrian levels of urban canopies, accurate validation with wind-tunnel experiments is essential. While extensive LES research has been conducted, validation often focuses solely on fundamental statistical profiles above the urban canopies, disregarding the dramatic spatial variations and higher-order statistics within the flow field, particularly at pedestrian levels. Therefore, this study systematically validated the spatial distribution of various LES-derived statistics against PIV data within a cubical array, focusing on critical height of $0.1H$ (where $H$ is the cube length). In addition to fundamental statistics, higher-order statistics were quantitatively validated with a hot-wire anemometer (HWA) above the canopy. Furthermore, to ensure the reliability of the LES results, a sensitivity analysis was conducted to assess the impact of the mesh resolutions and domain sizes. The $H/40$ mesh aligns with PIV results for mean and standard deviation at $0.1H$, offering a balance between accuracy and computational cost. In terms of validation metrics, the $H/60$ mesh indicated the best consistency with the velocity skewness and kurtosis obtained by HWA above the canopy. While the turbulent length scale is affected by domain size when using cyclic boundary conditions, utilizing a $16H$ of streamwise length reproduces a more reliable streamwise integral length scale with the experiment for $4H$ and $8H$ spanwise lengths. However, a sharp reduction in the streamwise integral length scale was observed when using $2H$ in spanwise length. By incorporating these detailed validations, this study aims to underscore the key roles of both mesh resolution and domain size in accurately verifying and validating LES models for simulating pedestrian-level winds.