Simple beats complex: Well-evaluated numerical urban climate models can provide simulations of canopy layer air temperature and urban heat islands in areas which lack observations. Many of these models, however, need significant computational resources, and the complex parameterizations involved require expertise to carry out the simulations. We have developed a simple statistical model which can predict the maximum urban heat island intensity under specific weather conditions based on the urban fraction and height-to-width ratio of a neighbourhood of interest and four weather variables measured at a reference location. Non-dimensional analysis is used to derive the semi-empirical equation. The approach works best when conditions for nighttime maximum heat island development are 'ideal’ (dry, calm and clear). Under these conditions, model-observation error metrics are better than those obtained using a more complicated numerical urban climate model.

The image on the right shows the mean nighttime urban heat island map for Singapore for 'UHI_maximum’ using the semi-empirical statistical model applied to 300 m gridded morphological and land cover data for ‘ideal' conditions. The highest individual grid-cell modelled UHI_max intensities reach 6.0 ºC (located in the CBD), and the most frequently occurring UHI_max values are between 4.5-5 ºC. The area-weighted UHI_max intensity considering all built-up areas on Singapore main island is 2.64 ºC - this can be interpreted as the nighttime city-wide average air temperature increment caused by the presence of the city under ‘ideal’ conditions.