Urbanization has been producing drastic changes in land surface characteristics. Cities become more densely built, and naturally vegetated surfaces are replaced with man-made asphalt and concrete. Increased surface roughness due to artificial buildings and structures leads to the reduction of wind speed and the increase in air temperature within urban canopy. Urbanization has significant impacts on the local environment that creates an "urban climate". Moreover, urbanization has a potential impact on severe local weather such as high wind and heavy rainfall, which raises awareness about the importance of urban weather. Recent societal issue in urban weather is related to the viewpoint of our social safety and security. In considering urban weather and climate, local flow and thermal structures within urban canopy should be investigated from a turbulence perspective.

Thus, many researchers have studied the turbulent flow over urban-like roughness models. For example, Macdonald (1998) proposed an improved model for the estimation of surface roughness of obstacle arrays. Cheng and Castro (2002) conducted a wind tunnel experiment and investigated the characteristics of the turbulent flows over a number of urban-type surfaces. Kanda (2006) performed large-eddy simulation (LES) on the turbulent flows of various surface geometries of building arrays and investigated the sensitivity of the drag coefficient to the surface geometry. However, the geometries of obstacles employed in these studies are too simplified to directly apply their results to real urban settings. On the other hand, the shape of city surfaces is complex and the building heights are highly variable. Ratti (2002) reported that the ratio of the standard deviation of building heights to the mean building height show 1.0 for some urban areas. Therefore, turbulent flow structures in realistic urban canopy should be further investigated.

In this study, we first examine the building morphological characteristics such as roughness density, the mean and standard deviation of building heights in actual urban area. From this analysis, we propose a model that represents realistic urban surface geometries. Next, we perform LES on the spatially-developing boundary layer flows over the above-mentioned building arrays and investigate the relationship between the turbulent flows and the building morphological characteristics.

The LES results for flows over uniform-height building arrays indicate that the turbulent properties agree well with the proposed models for the estimation of surface roughness due to roughness arrays (Raupach (1991), Macdonald (1998), Shao (2005)). However, the LES of flow over building arrays with variable heights shows that the turbulent characteristics largely differ from the previous models, especially in case of large roughness densities. This implies that it is difficult to apply the previous models based on the turbulence characteristics in roughness arrays with uniform heights to actual urban area.