Large-eddy simulation of the convective boundary layer over an idealized patchy urban surface

This study investigates the structure of thermal plumes driven by different patchy urban surfaces using large-eddy simulation. An idealized surface pattern with different sizes of 'park area' and 'built-up area' is assumed. These two types of urban surface correspond to 'colder' and 'hotter' areas over which different surface sensible-heat fluxes are specified. The stability parameter Z(i)/L (where Z(i) is the boundary-layer depth and L is the Obukhov length-scale) is large enough that a free-convection regime is considered. The study has three main aims: (i) to examine the effects of different surface patterns, including its 'skewness', on the turbulent structure and growth rate of an urban convective boundary layer (UCBL); (ii) to test the influence of model configuration, such as the domain size and the vertical resolution in the entrainment layer, on the turbulent structure of a UCBL; and (iii) to investigate the behaviour of thermal plumes in the UCBL when a weak wind shear is present. The results show that: (i) the surface pattern plays an important role in determining the power spectrum of the vertical-velocity fluctuations-the positive skewness of the surface heat-flux pattern enhances both the lifetime of the surface-driven modes and the skewness of the vertical-velocity fluctuations, and patchy surface patterns cause higher boundary-layer growth rates for a time t within the range 4t(*) < t < 16t(*) (where t(*) = Z(i)/w(*) and w(*) is the convective velocity-scale), but contribute little at later times; (ii) the domain size must be large enough (larger than 5Z(i)) in order to resolve correctly the surface-driven mode, and the vertical grid resolution in the entrainment layer has a small effect on the UCBL growth and on the turbulent statistics in the UCBL; and (iii) the wind is shown to have smoothing effects on the power spectrum, similar to the findings of Dornbrack and Schumann in their study of a convective boundary layer over a one-dimensional wavy terrain-for small values of the ratio d/Z(i) (where d is the length-scale of the surface pattern) this effect is very significant, but for large values the effect is reduced.