Computational Modelling Group

Wind direction effects on urban flows

1st October 2007
31st March 2011
Research Team
Jean Claus
Zheng-Tong Xie, Ian Castro

Mean pressure distribution on ground and building-like obstacles and mean velocity vectors in mid-height plan for a 45º wind.


Improving our understanding of air flows within cities is crucial for numerous fields such as pollution dispersion, parameterisation of weather forecasting and climate prediction models, wind loadings on buildings, pedestrian comfort and many more. Confronted with a wide variety of urban landscapes and in an effort to extract general characteristics of urban flows, recent research has been conducted on simple models of an urban canopy: e.g. staggered arrays of uniform or random height cuboids. In order to bridge the gap between this model and real urban canopies, more parameters need to be introduced. In this project we aim at clarifying the effects of different wind directions on the flow above and within the canopy.


We are particularly interested in the velocity and pressure distributions, the drag force on the buildings and the logarithmic part of the streamwise velocity profile. However, just as most environmental flows, urban air flows are turbulent – their velocity and pressure fluctuate chaotically in time and space. Numerical simulations of such flows are notoriously demanding in computing power, especially when some – the largest if not all – scales of turbulence need to be resolved. For this project, large eddy simulations (LES) were conducted using CD-adapco’s commercial code StarCD. Each simulation required from 8 to 48 processors running simultaneously from 250 to 650 hours (10 to 30 days 24/7) making Iridis3 an ideal computing resource.


The results of the numerical simulations compared very well with the wind tunnel experiments conducted in parallel, giving us confidence in the findings. Although the mean flow within the lower part of the canopy showed only weak dependence on the direction of the wind aloft, the variations of the surface drag and of the parameters of the logarithmic law – non-linear and non-monotonic with the variation of wind angle – highlighted the importance of taking the wind direction into account when studying urban flows.


This work is funded by the Natural Environment Research Council (NERC) and the National Centre for Atmospheric Science (NCAS).


Physical Systems and Engineering simulation: CFD, Turbulence

Socio-technological System simulation: Built Environment

Algorithms and computational methods: Finite volume

Simulation software: StarCD

Software Engineering Tools: Emacs, Vim

Programming languages and libraries: Fortran

Computational platforms: Iridis, Linux, Windows

Transdisciplinary tags: Scientific Computing