Experimental SetupThe experiments were carried out in a 16 m long, conventional type boundarylayer wind tunnel with a test section 1.5 m wide and 1 m high.  To obtain a suffi-cient geometrical resolution in flow and concentration measurements a modelscale of 1:200 was selected.  By means of a specific setup of turbulence genera-tors (spires) and roughness elements on the wind tunnel floor the lower part ofthe atmospheric boundary layer (up to 100 m height in full scale) was modelledin the test section of the wind tunnel.  Systematic boundary layer measurementsand an iterative improvement of the spire - roughness configuration enabled eventhe constant shear layer found in field data to be modelled up to 100 m heightfull scale.  Despite of the large model scale and the rather small cross-section ofthe tunnel, the parameters of the boundary layer flow agree well with field datafor a suburban roughness (Figure 1). The height of the model boundary layerwas exceeding the height of the model building by a factor of four. According tocommon guidelines for physical modelling this ensures that the wind tunnel re-sults are not affected by the limited height of the model boundary layer and thatsimilarity of the model results to field conditions can be achieved (VDI 3783). A 2D fibre-optic Laser-Doppler-Anemometer (~LDA, DANTEC®) with 500 mmfocal distance was used for non-intrusive flow measurements.  Due to the longfocal distance, flow disturbances caused by the LDA probe head were not de-tectable even for near-wall  measurements.  For concentration measurements, afast FID (flame ionisation detector, Cambustion Ltd.®) with a frequency resolu-tion of up to 200 Hz and a mobile sampling head was used.  Both, the LDA fi-bre-optic probe as well as the FastFID sampling unit were mounted on a com-puter controlled 3D traversing system to ensure sub-millimetre positioning accu-racy and high efficiency of automated measurements at more than 1000 meas-urement points per test case.  Besides mean values and statistical parameters ofthe measured quantities, time series of the flow velocity components and theconcentration of a tracer were recorded and analysed off-line with respect totheir fluctuation characteristics.  Two source configurations with different com-plexity have been studied in detail.  The first setup was consisting of an isolatedrectangular building with four area sources attached on the lee side of the build-ing (configuration A) close to the ground.  As second configuration, a finite ar-ray of 27 rectangular buildings with exactly the same shape and dimension wasinvestigated in the wind tunnel (configuration B).  In configuration B four rowsof model buildings with three buildings per row were mounted upstream of thesource building and two rows of buildings were attached downstream (Figure 2).3.  ResultsIn Figures 3 and 4, measurements of two components of the velocity field arepresented.  The results illustrate, that the average flow pattern is symmetrical forwinds perpendicular to the front side of the building(s).  For configuration A, thecharacteristic flow structures like flow separation on the flat roof of the buildingas well as on both sides of the model could be measured.  The typical horseshoe-vortex structure in front of the obstacle and the large recirculation area on theleeward side of the model building can be detected as well (Figure 3).  The sec-ond setup (configuration B) did not show flow separation on top of the modelbuildings.  In this configuration the separation zones on the sidewalls of thebuilding were replaced by an almost uniform flow in the street canyons parallelto the wind.  For the crosswind street canyon behind the source building anasymmetric recirculating flow pattern was detected in the vertical measurementplane which differs significantly from the almost centric flow pattern usuallyfound in a long street canyon (Figure 4).  The flow in configuration B was alsofound to be more unstable and intermittent than in configuration A. Long aver-aging times, corresponding to several hours in full scale, were required to cap-ture even the aperiodic changes of the flow patterns observed during LASERlight sheet visualization experiments and in order to get representative results. Example concentration distributions are given in Figure 5 and 6.  地下车库的污染物扩散英文文献和中文翻译(2):http://www.youerw.com/fanyi/lunwen_69929.html