Air temperature (Ta) is ubiquitous in empirical modelling of future glacier extent. However, we argue that this long-held convention should be overturned in favour of the equivalent temperature Te, because the latter is proportional to the total (sensible plus latent) atmospheric heat exchange with a moist surface. From polar to tropical regions we show that variations in the glacier surface energy balance are better captured using Te, with the largest improvement in the critical tropical water tower regions. Similarly, we find that Te can discriminate more skilfully between rain and snow at 90 % of 5509 locations assessed. We also find that on-glacier Te is easier to infer from spatially coarse climate data. For these reasons we conclude that there are very strong, theoretical, empirical, and practical motivations for a paradigm shift in replacing Ta with Te to simulate the impacts of rising atmospheric heat content on glacier mass balance.