In building performance simulations it is typically assumed that outdoor HVAC equipment is exposed to air at the outdoor ‘ambient’ temperature, obtained from a weather file. However, in reality, significant spatial variations in the outdoor air temperature can exist close to buildings. This simplification may lead to significant errors, especially when comparing simulations of buildings with and without cool roofs, since the colder surface of the cool roof would reduce above-roof air temperatures, thereby influencing the energy consumption of rooftop HVAC equipment. Above-roof temperature variations were measured in detail at three shopping centre buildings, and the experimental data was used to develop a model that can estimate air temperatures above hot or cold roof surfaces. The model requires only four input variables, all of which are available in typical building simulation tools. Implementation of the model in a set of case-study simulations revealed that above-roof air temperatures can have a large effect on the predicted performance of cool roofs. In cases where cooling equipment and ventilation inlet ducts were both located on the roof, the electricity savings and gas ‘penalties’ attributable to cool roofs would have been underestimated by 44–85% (61% on average) if above-roof air temperature variations had not been modelled accurately.