TracNav
- TOC "gallery/toc" is empty!
Urban neighbourhood LES
The following visualizations show different flow aspects and phenomena of the neutrally stratified urban boundary layer. Simulations differ in domain size and grid resolution.
Non-Cyclic Turbulent Inflow - Turbulent inflow versus laminar inflow for large-eddy simulations (2015)
Project: A new turbulent inflow method for PALM and the effect of tall buildings on the urban boundary layer
Responsible: Tobias Gronemeier?, Helge Knoop?
Description: The animation shows a comparison between two large-eddy simulations (LES) using different inflow boundary conditions. For both simulations the LES model PALM was used, simulating a neutral stratified flow over an array of building cubes. The upper half of this visualization shows a simulation, which uses a laminar inflow at the left boundary while the lower half shows a simulation, which uses a turbulence generator based on a filter method at the left boundary. The size of both domains is 2180m x 720m x 240m with a mean background wind of 6 m/s at the top of the domain blowing from left to right. The rotation of the velocity vector (absolute values) is visualized to show the turbulence structures and intensities. High values are marked red while low values are white. The buildings have a cubic shape with 24m edge length and are packed with a plane area index of 0.25. One tall building sits in the center of the domain with three times the size of a small building in horizontal direction and four times the size in vertical direction. The animation was created using the visualization software VAPOR. Simulations were calculated on the Cray-XC30 of the North-German Supercomputing Alliance (HLRN) as well as on the TSUBAME 2.5 of the Tokyo Institute of Technology (Tokyo Tech).
In the simulation with laminar inflow (top), first turbulent motions can be spotted behind the tenth building row. In reality such a laminar flow is almost never observed and hence very artificial. In the simulation with generated turbulent inflow (bottom), turbulence is created at the inflow boundary. This leads to an already turbulent flow above the first building rows. This flow is much more realistic. The flow in close vicinity to the tall building at the center of the domain shows slight differences between the two simulations. These differences can especially be seen at the rooftop of the tall building. Here the arriving flow in the top simulation shows almost no developed turbulence, while the arriving flow in the bottom simulation is already turbulent. At the outflow boundary however, both simulations show nearly equally developed turbulence.
The description above indicates, that the presented turbulence generation method allows legitimate analysis of simulation data much closer to the inflow boundary which can result in significant cost savings due to smaller required domain sizes.
| Model Setup | |
|---|---|
| Total domain size (x|y|z): | 2160m x 720m x 241m |
| Grid spacing (x|y|z): | 2m x 2m x 2m |
| Number of grid points (x|y|z): | 1080 x 360 x 122 |
| Simulated time: | 1.5h |
| CPU-time: | 5.5h |
| Number of CPUs: | 540 |
| Machine/ processor type: | TSUBAME 2.5 at Tokyo Tech and Cray-XC30 at HLRN |
| Visualization software: | VAPOR |
| DOI: | |
Influence of a densely build-up artificial island on the turbulent flow in the city of Macau (2014)
Project: High resolution large-eddy simulations of the urban canopy flow in Macau?
Responsible: Marius Keck?, Helge Knoop?
Description: The animation displays the development of turbulence structures induced by a densely built-up artificial island off the coast of Macau. Animation data were derived using the parallelized large-eddy simulation model PALM, simulating a neutrally stratified flow over Macau, with a mean flow from the southeast to the northwest and a 10-m wind of approximately 1m/s. The vertical direction of the model domain is stretched by a factor of 3 for better visualization. Turbulence structures and intensities are visualized by the rotation of the velocity vector (absolute values), with highest values in red and lowest values in white . Buildings are displayed in blue. The animation spans over 1 hour with a time-lapse factor of 43, and was created with the visualization software VAPOR. The total PALM model domain had a size of 768 x 256 x 96 grid points in streamwise, spanwise and vertical direction, with a uniform grid spacing of 8m in each direction . Above 400m the vertical grid spacing is successively stretched up to a maximum vertical grid spacing of 40m. Non-cyclic boundary conditions are used in streamwise direction and a turbulence recycling method is applied, in order to guarantee a fully turbulent inflow. In total, the simulation required 1 hour of CPU time using 128 cores on the Cray-XC30 of the North-German Supercomputing Alliance (HLRN).
The approaching flow above the sea shows a comparatively low turbulence intensity due to the smooth water surface. Within the building areas, strong turbulence is generated by two main reasons. One is the additional wind shear due to the walls of isolated highrise buildings. Furthermore, due to the significant increase in surface roughness, a so called internal boundary layer with enhanced turbulence develops above the building areas. The depth of this layer grows in downstream direction .
During the animation the camera moves through three major viewing angles . The first part of the animation starts with an aerial view onto the whole Macau area. Afterwards the camera zooms in, displaying those areas of the model domain , in which the flow field is particularly influenced by buildings. The second part is a side view from close above the surface and shows the above mentioned internal boundary layer. The last part shows another aerial view focusing on the gap between the artificial island and the Macau Peninsula, where turbulence decreases as it is advected across the gap.
| Model Setup | |
|---|---|
| Total domain size (x|y|z): | 6144m x 2048m x 768m |
| Grid spacing (x|y|z): | 8m x 8m x 8m |
| Number of grid points (x|y|z): | 768 x 256 x 96 |
| Simulated time: | 1h |
| CPU-time: | 1.5h |
| Number of CPUs: | 128 |
| Machine/ processor type: | Cray-XC30 at HLRN |
| Visualization software: | VAPOR |
| DOI: | 10.5446/14368 |
Turbulent flow around high-rise office buildings in downtown Tokyo (2008)
Project: High resolution LES of turbulent flow in the vicinity of buildings including thermal effects
Responsible: Marcus Letzel?
Description: The parallelized large-eddy simulation (LES) model PALM simulates a neutral turbulent urban boundary layer in Shinjuku, downtown Tokyo, Japan using GIS data provided by CADCENTER, Tokyo. The simulation lasts 3 h with a domain size of 900 m x 900 m x 492.5 m, periodic boundary conditions and 5 m uniform grid length. The model is driven by a 1 m/s westerly wind applied at the top of the domain (Couette flow) and initialized with a vertical profile obtained from a 1D model prerun. PALM’s Lagrangian particle model is used to track passive tracers with 1 h lifetime that are released every 5 min from four vertical line sources (colour ~ current height, tail length ~ velocity).
Particles in front of the metropolitan twin towers travel far upstream close to the ground during the first 15 min because turbulence has not yet fully developed. During most of the simulation the flow is channeled by tall buildings acting as street canyons. Zoom views show particle paths under the influence of eddies and helical vortex structures shed off the large buildings. Intermittent low-level upstream flow is evident in the most southerly street canyon particularly during t = 92…121 min when several blue (low-level) particles travel westwards (upstream). These features highlight the ability of LES models to capture turbulent fluctuations.
The Doc-Show Virtual Reality (DSVR) framework used for visualization consists of three subsystems: a parallelized library coupling to PALM, where geometry is created via FORTRAN or C function calls, a streaming server receiving the simulation results, storing the geometry and serving viewing clients and a browser plugin permitting real-time interactive presentation in various modes, including variation of lighting and thickness of particle tails. The level of detail of the geometry can be seen in wireframe mode during t = 62…80 min, when all rendered surfaces are shown as triangular primitives. The resulting transparency effect permits to look through surfaces.
Original online publication:
- Letzel, M. O., G. Gaus, S. Raasch, N. Jensen and M. Kanda, 2008: Turbulent flow around high-rise office buildings in downtown Tokyo. Dynamic Visualization in Science, No. 13100, originally published under the URL http://www.dyvis.org/DyVis?Sig=13100.
Associated journal publication with cross-reference:
- Letzel, M.O., M. Krane and S. Raasch, 2008: High resolution urban large-eddy simulation studies from street canyon to neighbourhood scale, Atmos. Env., 42, 8770-8784. doi:10.1016/j.atmosenv.2008.08.001.
| Model Setup | |
|---|---|
| Total domain size (x|y|z): | 900.0m x 900.0m x 492.5m |
| Grid spacing (x|y|z): | 5m x 5m x 5m |
| Number of grid points (x|y|z): | 180 x 180 x 100 |
| Simulated time: | 10800s |
| CPU-time: | 1.1h |
| Number of CPUs: | 30 |
| Machine/ processor type: | IBM-Regatta / Power4 |
| Visualization software: | DSVR |
Turbulent flow in a densely built-up area in Kowloon, downtown Hong Kong (2008)
Projects:
- LES of quarters in Hong Kong
- Turbulence structure in the urban roughness sublayer: LES reference studies and comparison with data from wind tunnel, scale model and field measurements
Responsible: Marcus Letzel?
Description: To follow.
Original online publication:
- Letzel, M. O., G. Gaus, 2008: Turbulent flow in a densely built-up area in Kowloon, downtown Hong Kong. Dynamic Visualization in Science, No. 13118, originally published under the URL http://www.dyvis.org/DyVis?Sig=13118.
Associated journal publication with cross-reference:
- Letzel, M.O., C. Helmke, E. Ng, X. An, A. Lai and S. Raasch, 2012: LES case study on pedestrian level ventilation in two neighbourhoods in Hong Kong. Meteorol. Z., 21, 575-589, doi: 10.1127/0941-2948/2012/0356.
| Model Setup | |
|---|---|
| Total domain size (x|y|z): | 768.0m x 512.0m x 400.0m |
| Grid spacing (x|y|z): | 1m x 1m x 0.8m |
| Number of grid points (x|y|z): | 768 x 512 x 325 |
| Simulated time: | 10800s |
| CPU-time: | 77h |
| Number of CPUs: | 128 |
| Machine/ processor type: | IBM-Regatta / Power4 |
| Visualization software: | DSVR |
LES of a city quarter of Hanover (Allianz tower, 2006)
Project: none
Responsible: Siegfried Raasch?
Description: Turbulent flow around a city quarter of Hannover. The mean flow is from west (right) with a speed of 1 m/s (neutral stratification is assumed). Clouds of particles are periodically released in front of the large building (Allianz tower) and in the courtyard of another complex of buildings. The particle color reflects the height above ground (red: high, blue: low). Topography data are from laser altimeter measurements (kindly provided by the Institut für Kartographie und Geoinformatik, Leibniz Universität Hannover). The resolution of the laser data allows to resolve cars on the street west of the Allianz building, staying there because of a red traffic light. The sequence shows that the turbulent flow within street canyons is highly variable so that flow directions may change completely within short times.
| Model Setup | |
|---|---|
| Total domain size (x|y|z): | 256m x 256m x 200m |
| Grid spacing (x|y|z): | 1m x 1m x 1m |
| Number of grid points (x|y|z): | 256 x 256 x 200 |
| Simulated time: | 1h |
| CPU-time: | 1800s |
| Number of CPUs: | 32 |
| Machine/ processor type: | IBM-Regatta / Power4 |
| Visualization software: | DSVR |
| Download | ||
|---|---|---|
| Video Format | File size | |
| Flash (.flv) | 8.3mb | Download |
| MPEG (.avi) | 8.3mb | Download |
