Forest canopies reduce shortwave radiation and increase longwave radiation reaching the underlying surface, compared with open areas, and thus influence rates at which forest snowpacks melt. The sub-canopy radiative environment can be highly heterogeneous, with temporal persistence depending on canopy structure and differing for shortwave and longwave fluxes, and this influences the rate at which snow-free ground emerges during snowmelt. Arrays of radiometers have been used to measure spatial variability in forest radiation, but such instruments are expensive and require regular attention in snowy environments. Hemispherical photography allows rapid collection of canopy structure data, and many software packages have been developed for modelling transmission of shortwave radiation using hemispherical photographs, but modelling of longwave radiation has received much less attention. Results are used here from radiometers located beneath lodgepole pine stands of varying density at the Marmot Creek Research Basin in Alberta, Canada. A simple model using sky view calculated from hemispherical photographs to weight longwave emissions from the canopy, calculated using measured air temperature as a proxy for canopy temperature, and measured above-canopy longwave radiation is found to give good estimates for spatial averages of sub-canopy longwave radiation, although standard deviations are generally underestimated. If above-canopy longwave radiation is parametrized as a function of air temperature and humidity rather than measured, good results are still obtained for daily and longer averages of sub-canopy longwave radiation. A multiple linear regression model using measurements of above-canopy shortwave radiation to estimate daytime canopy heating gives better results in comparison with individual radiometers. Copyright © 2008 John Wiley & Sons, Ltd.