The urban environment quality and citizen's comfort are strongly affected by human activities and urban morphology. Based on the close connection between the urban canopy flow and the atmospheric surface layer, it is very important to understand exchange processes happening above and within the urban canopy.

Increasing computational power and using more accurate measurement tools deepen the understanding of atmospheric turbulence. However, turbulent mechanisms involved in urban canopy flows are very complex and their fine description and understanding remain an important scientific challenge.

The dynamic properties of the urban flows were studied in literature using numerical simulation such as direct numerical simulations (DNS) and large-eddy simulations (LES) where the urban canopy was modeled by cubical obstacles arrays of varying density. Such simulations are demanding in terms of computational resources, especially for a large domain.

For the study of real urban areas, the detailed knowledge of the buildings organization is actually unavailable and a drag-porosity approach may be preferable, as commonly used for forest canopies. The drag-porosity approach models the presence of obstacles and their influence on the turbulent flow by a drag force which depends on averaged morphological characteristics of the canopy. It has the advantage of reducing the computation costs and can be applied for example to generate realistic inflow conditions for studies at the scale of some buildings. However, only few works exist on the use and efficiency of drag-porosity models to represent the turbulent transfers between urban canopies and atmosphere.

In this paper, high order statistics, turbulent coherent structures and turbulent kinetic energy budget are computed from an obstacle-resolved simulation and a drag-porosity model simulation. The objective of this paper is to highlight the capacity of the drag-porosity approach to reproduce unsteady turbulent flow over an urban canopy and to evaluate in details its performance.

Large eddy simulations of the same urban-like configuration are performed with the open source software OpenFOAM either using a drag-porosity approach or resolving obstacles. The urban canopy is represented by an array of staggered cubical obstacles with a 25% packing density. The Reynolds number based on the top velocity and cube height (h) is 5000. The computational domain is Lx * Ly * Lz = 16h * 12h * 8h. Periodic boundary conditions are imposed in horizontal directions, a free slip condition is set on the top of the domain and no-slip boundary conditions are applied on the bottom of the domain and on all obstacle surfaces for the resolved simulation. The resolved domain has around 22 million cells with a minimum resolution of h/32 around the cubes whereas the porous domain as 1.6 million cells with a minimum cell size of h/8.

High order statistics, turbulent coherent structures and turbulent kinetic energy budget are computed from instantaneous results of obstacle-resolved simulation and drag-porosity model simulation. First results show that the mean velocity profiles obtained by the two approaches are in good agreement to each other and to the data from literature. Higher order statistics are also comparable, but differences are observed in several variables such as the turbulent length scales. This work aims at describing and analyzing precisely the differences observed in turbulence characteristics that may be responsible for the transfers between urban canopy and atmosphere.