Revisit of Wind Flows between Wind Tunnels and Real Canyons – The Viewpoint of Reynolds Dynamic Similarity

Abstract

Wind tunnel test and numerical simulation are two powerful methods to study air flows and pollutant dispersions around urban buildings environment. As commonly known, the development of a successful numerical model should be firstly validated by experimental results, usually by wind tunnel data [1,2], a numerical model is ultimately needed to simulate the air flows and pollutant dispersions in real street canyons [3]. It is obvious that wind tunnel models and real canyons are of different scales. The scaling effects for wind tunnel tests have already been investigated on some simple models, such as single building and normal street canyon with the aspect ratio near 1.0 (AR, denoted by building height, H, over road width, W) [4,5]. But the scaling effects on the flows around complex buildings or in real street canyons still need to be examined. In wind tunnel test, air is often used as the experimental fluid. The air velocities in the wind tunnels are several metres per second, rarely more than 20m s 1 [5,6], which are similar to the velocities in the real canyon environment. But wood blocks with height of several centimetres, subject to the experimental conditions, are adopted for the street canyon models, following that the dimensions of wood blocks are about two orders of magnitudes smaller than the real buildings. Thus, the wind tunnel experiments only satisfy the geometric similarity, but miss the Reynolds dynamic similarity in nature. On the other hand, the numerical simulations of turbulent flows with high Reynolds numbers often demand large computing resources, so that the downscaling model is always used to simulate the air flows and pollutant dispersions in urban street canyons. In such cases, building models are set in several centimetres and incoming wind velocities are given similar to the real background wind velocities; this treatment is the same as the wind tunnel test [7,8]. This means that the downscaling simulations would just satisfy the geometric similarity, but miss the Reynolds dynamic similarity. The downscaling modelling was based on the Townsend’s ‘‘Reynolds number similarity’’ hypothesis [9] and determined by the critical Reynolds numbers [1,10]. The Townsend’s hypothesis is that, in the absence of thermal and Coriolis effects and for a specified flow system whose boundary conditions are expressed non-dimensionally in terms of a characteristic length, L, and velocity, UR, the flow structure is similar with all sufficiently high Reynolds numbers. Based on the ‘‘Reynolds number similarity’’ hypothesis, the critical Reynolds numbers are only determined by the representation of minor changes in flow structures or wind profiles, in which actually the wind