Street Canyon Ventilation of Traffic Pollution: Comparison Between Experiment and LES Kellnerová, Radka * 1,2) , Fuka, Vladimír 2 , Kukačka, Libor 1,2) , Uruba, Václav 1) , Nosek, Štěpán 1) , Jaňour, Zbyněk 1) 1) Institute of Thermomechanics, Czech Academy of Sciences, Prague, Czech Republic 2) Department of Meteorology and Environment Protection, Charles University, Prague, Czech Republic *)radka.kellnerova@it.cas.cz ABSTRACT The paper deals with comparison between experimental data and a numerical computation with LES simulation of a velocity, a momentum flux and a concentration flux in a street canyon. 1 INTRODUCTION According to Hertwig et al. (2011), the feasible strategy for a validation of large eddy simulation (LES) is not only the mean-values comparison since there might not be sufficiently representative data in terms of statistics neither from LES nor from time-resolved particle image velocimetry (TR-PIV). The more suitable approach could be some of the temporal-spatial analyses such as Fourier/Wavelet analysis, correlation or Proper orthogonal decomposition (POD). As suggested by Jimenez (1998), momentum flux is also a sensitive quantity for testing the proper LES equations closure. This work compares the forecasts of flow dynamics from two versions of LES with measured data by means of both approaches. 2 SET-UP AND RESULTS 2.1 Experiment Experiment is conducted in the wind channel having its floor covered by an array of street canyons. Two shapes of the building roofs are used – a triangular and a flat one. In this paper, results from the pitched version (Figure 1) are published. The aspect ratio of the street canyon is 1. The more detailed description of the set-up can be found in Kellnerova et al. (2011). The longitudinal, lateral and vertical direction corresponds to X, Y, and Z-direction, respectively. The 2D velocity information is obtained by TR-PIV measurements in the vertical plane (XZ). The overview of the PIV system is listed in the table 1. The PIV images are post-processed by Adaptive correlation software with final interrogation area 32 x 32 pxs. The spurious vectors were removed by Coherence filter and replaced by estimated median value. Table 1: PIV system Diode pumped Nd:YLF Repetition 500 Hz Camera resolution 1280 x 1024 pxs Interrogation area 32 x 32 pxs Overlapping 50% (80 x 64 vectors) Spatial resolution 1.2 mm x 1.2 mm Illuminated Area 100 x 100 mm Acquisition time 3.2 s