In this paper, we simulate temperature reductions during heat-wave events and during typical summer conditions from the installation of highly reflective "cool" roofs in the Chinese megacity of Guangzhou. We simulate temperature reductions during six of the strongest historical heat-wave events over the past decade, finding average urban midday temperature reductions of 1.2 °C. In comparison, we simulate 25 typical summer weeks between 2004 and 2008, finding average urban midday temperature reductions of 0.8 °C, indicating that air temperature sensitivity to urban albedo in Guangzhou varies with meteorological conditions. We find that roughly three-fourths of the variance in air temperature reductions across all episodes can be accounted for by a linear regression, including only three basic properties related to the meteorological conditions: mean daytime temperature, humidity, and ventilation to the greater Guangzhou urban area. While these results highlight the potential for cool roofs to mitigate peak temperatures during heat waves, the temperature reductions reported here are based on the upper bound case, which increases albedos of all roofs (but does not modify road albedo or wall albedo).