The implementation of ‘cool’ roofing materials, with high solar reflectance and infrared emittance, has received significant attention in recent years, as a method to mitigate the urban heat island effect and reduce building cooling energy requirements. The effect of ‘cool’ roofs on heat transfer through the roof structure has been investigated by many researchers. However, the air temperature field above roofs and the influence of elevated above-roof air temperatures on the performance of rooftop air-conditioning equipment and photovoltaic panels have not been studied in depth.
This paper describes detailed measurements that were taken in the thermal boundary layer above a large-footprint building. Air and roof surface temperatures were monitored at up to 63 locations above the roof of the building, and a comprehensive set of weather parameters were logged. Additional measurements were taken on specific days to characterise the atmospheric boundary layer profiles upwind of the building, measure air velocities above the roof, and determine the roof surface solar reflectance and infrared emittance. Roof surface temperatures were observed to often exceed the ambient air temperature by 30°C during the middle of the day. Air temperatures within 1.5m of the roof surface were typically 0.5-4°C above the ambient air temperature for several hours each day.
Computational Fluid Dynamics (CFD) simulation of convective heat transfer from building roofs is challenging, due to the complex urban geometries and high Reynolds numbers involved, and the wide range of relevant Richardson numbers, which span forced, mixed and natural convective regimes. Accurate simulation of thermal and velocity boundary layers is essential for such cases, but extremely fine computational grids are required to do so without the use of wall functions.
The ability of CFD simulations using wall functions to accurately model above-roof temperature fields was tested in the present work. Four CFD methods were compared, including delayed detached eddy simulations and wall-modelled large eddy simulations. Simulations of two cases, characterised by natural and mixed convection, were compared with results from the experimental campaign. The RMS deviation between simulated and measured air temperatures near the roof surface ranged from 0.74°C to 2.26°C. These discrepancies were of the same order of magnitude as the temperature differences that were of interest, which indicated that CFD methods involving wall functions may not be suitable tools for the investigation of above-roof temperature fields.