AbstractWind-tunnel experiments were carried out on four urban morphologies: two tall canopies with uniform height and two super-tall canopies with a large variation in element heights (where the maximum element height is more than double the average canopy height, $$h_{max}=2.5h_{avg}$$
h
max
=
2.5
h
avg
). The average canopy height and packing density are fixed across the surfaces to $$h_{avg} = 80~\hbox {mm}$$
h
avg
=
80
mm
, and $$\lambda _{p} = 0.44$$
λ
p
=
0.44
, respectively. A combination of laser Doppler anemometry and direct-drag measurements are used to calculate and scale the mean velocity profiles with the boundary-layer depth $$\delta $$
δ
. In the uniform-height experiment, the high packing density results in a ‘skimming flow’ regime with very little flow penetration into the canopy. This leads to a surprisingly shallow roughness sublayer (depth $$\approx 1.15h_{avg}$$
≈
1.15
h
avg
), and a well-defined inertial sublayer above it. In the heterogeneous-height canopies, despite the same packing density and average height, the flow features are significantly different. The height heterogeneity enhances mixing, thus encouraging deep flow penetration into the canopy. A deeper roughness sublayer is found to exist extending up to just above the tallest element height (corresponding to $$z/h_{avg} = 2.85$$
z
/
h
avg
=
2.85
), which is found to be the dominant length scale controlling the flow behaviour. Results point toward the existence of a constant-stress layer for all surfaces considered herein despite the severity of the surface roughness ($$\delta /h_{avg} = 3 - 6.25$$
δ
/
h
avg
=
3
-
6.25
). This contrasts with the previous literature.