Effects of minor changes in the mean inlet wind direction on urban flow simulations

There has been increased prevalence in the use of computational fluid dynamics (CFD) for the assessment and planning of the urban environment in recent years, spurred not only by the decreasing costs of computational resources, but also by increasing demand for enhanced living conditions, especially in densely populated and land-scarce places like Singapore. In spite of this, there are still many unknowns as to how the simulation parameters affect the prediction accuracy. In the present work, an assessment is carried out to investigate how small modifications to the mean inlet wind direction affect the simulation results. Field measurements performed within a town in Singapore are compared to three-dimensional (3D) steady Reynolds-averaged Navier-Stokes (RANS) simulations for seven inlet wind directions within a 30 degrees sector, using the realizable k-epsilon model. The driving conditions for the simulations, i.e. the overlying wind conditions, are derived from nearby meteorological station data, which is also used in the sampling of the field measurements. The simulation results compare favourably with the field measurements for the wind speed, while the veracity of the computed wind directions is highly dependent on the immediate surroundings of the sensors. Additionally, major changes in the simulated flow patterns are observed even for small changes in the mean inlet wind direction. A single simulation result cannot be considered representative of a specific mean inlet wind direction - this is evidenced by high variation in the field measurements even after stringent sampling criteria. The results have implication particularly in the field of urban planning, and highlight the care that should be taken when CFD results are used as part of the decision-making process.