A dynamic urban canopy parameterization based on CFD microscale modeling

One of the peculiarities of the atmosphere in the Urban Canopy Layer is that the structure of the mean fields of the meteorological variables (wind, temperature, etc.) presents strong heterogeneities at the same scale of the size of the elements of the urban morphology (e. g. few meters). This is different than what happens over flat and homogeneous terrain, where, even if there are turbulent structures at all scales, the mean fields vary only with height. Another important spatial scale for the urban atmosphere is the scale of the whole city. It is at this scale (of several tenths of kilometers, or mesoscale) that the circulations induced by the differences between urban and rural landuses form. For many applications (air quality, urban climate, etc.), both scales are important and should be considered in the modeling process. However, a domain containing the whole city and its surroundings can not have a resolution high enough to resolve urban obstacles explicitly. It is necessary then to cut the range of scales explicitely simulated and parameterize the effects of the scales smaller than the grid size of the model (typically several hundreds of meters to few kilometers). The outputs of these Urban Canopy Parameterizations (UCP) are then spatial averages over the grid cell of the model (usually a mesoscale model). However, it is often problematic to get these spatially averaged variables because due to the hetereogenites mentioned above, point measurements inside urban canopy layer have a limited spatial representativeness, and they are not representative of the spatially averaged variables. In this contribution a Computational Fluid Dynamic (CFD) model, that can be run over a domain that do not cover the whole city but resolve explicitly the flow around buildings, is used to design, calibrate and validate UCPs. The advantage is that spatial averages can be computed from CFD results and compared with UCP outputs. In this study, the use of Reynolds-averaged Navier-Stokes (RANS) – CFD simulations over simple geometries (staggered arrays of cubes) is twofold. Firstly, UCP inputs (length scales and drag coefficient) are calculated using RANS – CFD results and simple parameterization for input variables are proposed. Secondly, UCP with different proposal of drag parameterizations are evaluated by a comparison between UCP output and spatially averaged variables from RANS – CFD results.