Ventilation of urban heat islands is a major motivation for bulk transfer coefficient studies involving heated urban surfaces using field measurements, wind tunnel experiments, and numerical simulations. Dependence of these coefficients on numerous variables, such as flow structure, thermal stability, domain geometry and scale, and reference position, complicates intercomparisons for field and wind tunnel experiments. Numerical simulation, specifically Large Eddy Simulation (LES), has the advantage of normalizing extraneous variables to enable focus on the turbulence effects of a single determining factor, and it does this for a three-dimensional, sufficiently large urban domain with an appropriately fine resolution.

The purpose of the study is to compare the effect of each urban surface on heat ventilation of an array of urban-like roughness by numerically computing bulk transfer and correlation coefficients. A constant heat flux is applied for each surface, whose position is relative to the mean wind direction – roof, ground, and windward, leeward, and streamwise walls.

Using the Parallelized LES Model (PALM) developed by the Institute of Meteorology and Climatology, University of Hannover, a 1 m/s wind is simulated over explicitly resolved cube arrays with a size of 24 m along one side, in a domain of 768 m along the streamwise & spanwise directions and 384 m high, and a constant heat flux of 0.1 K m/s for a surface. Cube area density of 1/4, and aspect ratio (cube height per canyon width, H/W) of 1, are used. Numerical domain size is 512 x 512 x 128 grids for a resolution of 1.5 m/grid, or 16 grids per cube length. Integration time is set at 12 hours, since quasi-steady flow is attained in less than that time for all cases. Deardorff-modified Smagorinsky Model is used to solve for subgrid-scale (SGS) turbulence, and the Temperton algorithm for Fast Fourier Transform to solve the Poisson equation of pressure. Boundary conditions include: zero initial gradient for velocity and temperature, cyclic for lateral, non-slip for bottom and slip for top.

The proximity, orientation, and horizontal projection of heating for each urban surface factor into momentum and heat exchange. Canyon surfaces have stronger canopy mixing, and parallel surfaces have faster mean winds and overall vertical fluctuations. Horizontal surfaces have the largest momentum fluxes and temperature fluctuations near their proximity, while vertical surfaces have more efficient turbulent fluxes. The roof case has the largest coefficients, followed by the windward wall, leeward wall, streamwise wall, and ground.