Applicability of two computational fluid dynamics models and one mesoscale model for predicting urban heat island circulation

Predicting city-scale air flows such as urban heat island circulations (UHIC) is challenging following the interaction of microscale winds (buildings, neighbourhoods and districts), and mesoscale winds beyond a city. Being at an intermediate scale, UHIC has been studied using both microscale prediction tools, e.g. computational fluid dynamics (CFD) and mesoscale weather forecasting tools, e.g. weather research and forecasting (WRF). A thorough comparison of the UHIC simulation capacities of these two classes of models has not been performed so far. In addition, current mesoscale models generally cannot resolve building-specific details or simulate fine-scale turbulence structures. Consequently, microscale models, such as the Parallelized Large-Eddy Simulation Model (PALM) and ANSYS Fluent, have gained prominence, as they can simulate building-resolved features. This study evaluates the trade-offs associated with the above-mentioned mesoscale and microscale models for a UHIC above an idealized square city with known complex flow patterns. The results show that WRF is computationally efficient and has a computational cost approximately one order of magnitude less than that of microscale models. However, WRF exhibits lower accuracy, particularly in capturing localized effects at a ground level. Fluent and PALM offer higher accuracy by simulating finer details at a higher computational cost than WRF. PALM and Fluent also exhibit similar accuracy, and PALM is more computationally efficient than Fluent. However, unlike Fluent, the use of Cartesian grids with PALM limits its ability in handling complex building shapes. Our results are useful for selecting the suitable city-scale prediction tools considering both computational costs and accuracy.This article is part of the theme issue 'Urban heat spreading above and below ground'.