We conducted Large eddy simulations (LES) of internal boundary layer developments over a huge urban area, i.e., 19.6 km in stream-wise and 4.8 km in span-wise, with 2m spatial resolution. The domain is a densely built-up urban area in Tokyo coastal region, including the Pacific Ocean at the inflow boundary. All the individual buildings without vegetation are explicitly resolved thanks to its high spatial resolution. A new LES technology using the lattice Boltzmann method has advantages in utilizing parallelized GPU computations, and contributes to realize such huge and detailed turbulence analysis of realistic urban boundary layers. The spatial development of internal boundary layer along the fetch from the coastal line are simulated and compared with the previous theories. Very large scale motions of longitudinally-enlarged streaky pattern are visible everywhere within the internal boundary layer, and increase the length ( a few kilometers) and distance (several hundred meters) in accordance with the fetch. The similar streaky patterns are observed by a Doppler rider conducted in the same region. High buildings or building clusters are likely to trigger streaks.

The wind flow over an urban area induced the gusty condition due to its heterogeneous landscape. It might lead to the unpleasant events either towards the pedestrian or any structure within the area. It motivates a study to represent the situation by an indicator called gust index. Therefore, a feasible and realistic large eddy simulation (LES) paired with the lattice Boltzmann method (LBM) over a 2 m resolution of a huge urban domain of Tokyo was conducted. This paper also proposed an appropriate definition of the gust index that characterized the level of gustiness of the urban area. The gust index was calculated by normalizing the maximum wind speed, umax in 10 minutes time duration to the freestream velocity, U∞. An understandable gust index map was prepared for an urban area in Tokyo. It can be referred to point out the area that creates the high gust occurrence. The potential factors such as the ratio of the open space to the building plan area, building height and size that contribute to the high gust index also will be discussed.

Key Words : Gust index, urban area, huge domain, LES, LBM

Understanding the turbulent transport of both momentum and scalars in urban environments is important for urban climatologists, boundary-layer meteorologists, and fluid dynamicists. Over such very rough surfaces caused in a large part by the buildings, the transport of momentum and scalars are dissimilar - form drag controls momentum exchanges with the surface while viscous exchanges continue to dominate for scalars. A numerical study is carried out using large-eddy simulations to investigate this dissimilarity. The development of key components of the code used in this study, and its evaluation will be presented in detail. Geometric parameters such as the frontal area index and the plan area index are then varied independently to examine their impact on turbulence and transport characteristics. The surface complexity is shown not only to increase the anisotropy of the flow, but also to modulate the efficiencies of momentum and scalar transport. Results suggest that distinctions between turbulent momentum and scalar transport make the parameterization of surface-air exchange over urban areas a non-trivial problem.

In most attempts conducted to couple engineering computational fluid dynamics (CFD) models with mesoscale meteorological model (MMM), RANS approaches have been selected as turbulence models. Recently, several attempts to couple large-eddy simulation (LES) and MMM have already been reported due to growth of computer resources. In order to couple LES with MMM successfully, One of the biggest unresolved issues is the generation of inflow turbulencewhich satisfies not only turbulent statistics but also instantaneous turbulent structures. For this purpose, several techniques have been developed in the fields of wind engineering and building science. These pioneering techniques to generate inflow turbulence within neutral boundary layer are classified into two categories. The first, and also the simplest, is to store the time history of velocity fluctuations given from a preliminary recycling LES computation (e.g. Lund et al., 1998; Kataoka and Mizuno, 2002). The second is to generate inflow turbulence artificially which prescribes turbulent statistics without LES (e.g. Lee et al., 1992; Iizuka et al., 1999; Klein et al., 2003; Xie and Castro, 2008; Xuan and Iizuka 2013). In recent years, the non-isothermal LES within urban boundary layer have been carried out to investigated heat island phenomena and so on. When LES is applied to a non-isothermal field, not only inflow velocity fluctuation but also temperature fluctuation should be reproduced. About the generation of inflow turbulence considering temperature fluctuation, only a few studies have been conducted based on the method by using the recycling LES which mentioned above (e.g. Tamura et al., 2012; Jiang et al. 2012) in wind engineering field.

This paper aims to propose a new method to generate turbulent fluctuations of wind velocity and scalar quantities such as temperature and pollutant based on the Cholesky decomposition of time-averaged turbulent fluxes tensor of momentum and scalar. The artificially generated turbulent fluctuations satisfy not only the prescribed profiles of turbulent fluxes of wind and scalar but also the prescribed spatial and time correlations. Following the method proposed by Xie and Castro (2008), two-dimensional random data are filtered to generate a set of two-dimensional data with the prescribed spatial correlation. Then, these data are combined with those from previous time step by using two weighting factors based on an exponential function. The method was validated by applying it to a LES computation of contaminant dispersion in a half channel flow.

The knowledge of external convective heat transfer coefficients (CHTC) is of great importance for many engineering applications. This is the case for building applications for example, where an accurate evaluation of CHTC is needed to calculate convective heat gains and losses from building façades and roofs. It is also important to evaluate the thermal performance of double-skin façades, solar collectors and greenhouses. Assessment of the CHTC can be performed by full-scale measurements, wind-tunnel experiments or Computational Fluid Dynamics (CFD) simulations. When CFD is considered, it can be a useful tool to determine CHTC in urban areas. In these cases, Large Eddy Simulation (LES) should be used because it can capture the complexity of the flow pattern much more accurately than Reynolds-Averaged Navier-Stokes (RANS) simulations. However, generating high-resolution high-quality computational grids for LES simulations of the CHTC is not straightforward. Ideally, LES grids, where the grid size is used as a filter, should consist of cubical cells. In addition, grids for assessing CHTC should have a very high resolution near the building surface in order to fully resolve the thin laminar sublayer (often only a few mm in thickness) which represents the largest resistance to convective heat transfer. In order to avoid a prohibitively high total number of cells and the need for excessive computational resources, we can resort to the development of a grid, consisting of cubical cells, that is refined gradually in all directions, i.e. a non-conformal grid. Nevertheless, the accuracy of this type of grid for the case of buildings in the highly turbulent atmospheric boundary layer still needs to be verified. Therefore, in this paper, the performance of non-conformal grids is evaluated. The evaluation is based on verification by a comparison between conformal and non-conformal grids and on validation with wind-tunnel measurements of surface temperature of a reduced-scale wall-mounted cube. The resulting number of cells for the conformal and non-conformal grid is 9.710.472 and 1.431.789 respectively, where both grids have the same near-wall resolution (y+ value). The results obtained by the non-conformal grid are in a very good agreement with the results by the conformal grid. In this case, the average difference between simulated surface temperatures on all surfaces of the cube is about 0.9% Moreover, the general agreement between the experimental results and CFD results using non-conformal grid is very good. For example, the average and maximum absolute deviation in surface temperature from the experiments are 2.0% and 5.5%, respectively. The verification and validation studies verify the accuracy of non-conformal grid for the investigation of CHTC for a wall-mounted obstacle. Finally, this means that the use of the non-conformal grid reduces the total number of cells by a factor 6.8, without compromising accuracy. To the best of our knowledge, LES CFD simulations of CHTC for a full-scale building have not been performed. In the second part of the paper, LES simulations of CHTC at the surfaces of a full-scale building are performed using a non-conformal grid. Further information concerning boundary conditions, solver settings and results will be presented in the full paper.

In order to accurately predict the wind flow in canopy layer of large urban area, we introduce LES(Large eddy simulation) based on BCM(Building Cube Method) which is formulated on very fine Cartesian mesh system. Houses and buildings were not modelled and directly reproduced their shapes, because the wind profile parameterization requires the correct estimation of local flow field in the canopy layer close to the ground. Recent high-performance computing (HPC) technique has developed distinctly, so high-resolution computation becomes able to be applied to flows around a complicated configuration such as actual urban area. In this case we have to deal with buildings, vegetation and street etc. as a part of numerical model. Actually LES using the Cartesian coordinate encounters the incorrespondence of directions between the street lines and the discretized mesh lines. Very fine mesh system by BCM can solve this problen supported by the external forcing technque at the boundary named IBM(Immersed Boundary Method). Also, in this numerical scheme, computational process is so simple that the parallel algorithm and the memory access obtain the perfect efficiency. It is strongly expected that these advantages make it possible to efficiently simulate the flow around very complicated shapes with various scales consisting of a large variety of urban parts. In this study, we have applied LES by BCM to the wind flow estimation over the real complicated urban surface at straightforward and inclined wind directions to the main streets. Computational domain is several kim square with resolution of 1m for urban area. We also have exhibited the database of the wind turbulence in canopy layer among the buildings.

We examine the numerical model for the BCM and 3D contouring surfaces of Q values. We can recognize the 3D vortical structures in the wake of buildings. Conical vortex is recognized to be reproduced on the roof clearly. These local flow structures are expected to provide the appropriate turbulent characteristics in urban canopy layer. It can be found that, based on the present LES results using BCM, the turbulence structures at inflow and among a pack of tall buildings are effective in comparison with the other data obtained by experiment. We can confirm that the BCM model estimates the wind flow above and within urban canopy with sufficient accuracy even at inclined wind direction to the main streets.

Using the HPC database for LES of turbulent boundary layer flow over urban-like roughened surface, analysis of turbulent wind in urban canopy is performed. Actual buildings and houses, definitely larger effective height (height to boundary layer thickness) than those on conventional rough wall, are directly arrayed on the ground surface and above- and within-region equilibrium profiles of mean velocity and turbulent statistics are investigated. Also, the turbulence structures in the urban canopy are elucidated and urban canopy parameterization is discussed.