Abstract
Urban forms characterised by multi-scale features in buildings can drastically modify
the wind structure within cities, affecting both pedestrian comfort and air quality at street
level. For simplicity, most urban flow studies focus on cuboid buildings with a single
length scale. In this work, wind tunnel experiments are conducted on six forms to assess
how additional length scales impact urban flow: two reference cuboid cases (Standard
and Tall) that differ in aspect ratio (mean building height to width), plus two additional
fractal iterations of each geometry. The six models have the same mean building width,
height, frontal and projected plan area but their length scale characteristics differ. When
considering the building models in isolation, the length scale differences are found to affect
the drag force exerted by the buildings in isolation in a non-negligible way (up to 5% and
13% for Standard and Tall buildings, respectively). The added length scales also modify
the wake lateral spread and intensity of the turbulence fluctuations within the wake flow,
with the smaller length scales having the lower (higher) intensity of fluctuations in the
near (far) wake. Additionally, the strength of the vortex shedding emanating from the
buildings is reduced by introducing systematically smaller length scales.
When the multiscale models are arranged in large arrays, with fixed frontal and plan
solidities, results show that the boundary layer height and roughness aerodynamic parameters
experience significant changes across fractal iterations. Changes are recorded in
both roughness and inertial sublayers’ depths with the addition of smaller model length
scales. Dispersive stresses are found to be up to 20% of the overall momentum flux for base
models, however these reduce significantly with fractal iteration. Quadrant and spectral
analysis show that the additional length scales act to dampen turbulent mixing below
the canopy layer, while enhancing it above the canopy height. The internal boundary
layer that develops over the multiscale arrays is examined, with changes to the growth
rate revealed when additional length scale are introduced. Furthermore, the addition of
smaller length scales significantly increases the fetch required for the roughness length
to fully adjust to the secondary roughness. Finally, different multiscale arrays are also
arranged in combination to explore how a change in roughness characteristics at different
fetch lengths of different length scales affects the internal boundary layer.