Broadband radiation schemes (parameterizations) are commonly used tools in glacier mass-balance modelling, but their performance at high altitude in the tropics has not been evaluated in detail. Here we take advantage of a high-quality 2 year record of global radiation (G ) and incoming longwave radiation (L ↓) measured on Kersten Glacier, Kilimanjaro, East Africa, at 5873 m a.s.l., to optimize parameterizations of G and L ↓. We show that the two radiation terms can be related by an effective cloud-cover fraction n
eff , so G or L ↓ can be modelled based on n
eff derived from measured L ↓ or G, respectively. At n
eff = 1, G is reduced to 35% of clear-sky G, and L ↓ increases by 45–65% (depending on altitude) relative to clear-sky L ↓. Validation for a 1 year dataset of G and L ↓ obtained at 4850 m on Glaciar Artesonraju, Peruvian Andes, yields a satisfactory performance of the radiation scheme. Whether this performance is acceptable for mass-balance studies of tropical glaciers is explored by applying the data from Glaciar Artesonraju to a physically based mass-balance model, which requires, among others, G and L ↓ as forcing variables. Uncertainties in modelled mass balance introduced by the radiation parameterizations do not exceed those that can be caused by errors in the radiation measurements. Hence, this paper provides a tool for inclusion in spatially distributed mass-balance modelling of tropical glaciers and/or extension of radiation data when only G or L ↓ is measured.