Average flow properties useful for urban parameterizations deduced from CFD simulations over a regular array of cubes

The study of pollutant dispersion at city scale (several tenths of kilometers) is becoming more and more important, both for ‘traditional' pollution (from cars, houses, industries), as well as for sudden releases of toxic gases (accident or terrorist attack). Being impossible to run a numerical model at such high resolution to resolve every building and cover at the same time the whole city, it is necessary to use non-obstacle resolving numerical approaches. One of this consists in using mesoscale models at very high resolution (less than one kilometer), and parameterize the effect of the buildings on the grid averaged variables. Several approaches have been proposed, mainly inspired to vegetation canopies modeling, but a thorough validation of such approaches is still missing. In this contribution, the results of the CFD model run over a regular array of cubes (presented and validated in Santiago et al., same session) are spatially averaged to deduce some information on the behavior of the mean and subgrid variables, which are the relevant variables for the mesoscale model. In particular, vertical profiles of mean wind, turbulent and dispersive fluxes, and building drag are analyzed. Furthermore, the CFD results will be used to deduce turbulent length scales and to estimate the subgrid variability of different variables. From such analysis several important information to test and improve urban parameterizations will be deduced. Among the others: the importance of the dispersive fluxes within the urban canopy, simple rules for the vertical profiles of the mean horizontal wind, and a modification of the drag parameterization.