Data collected at three Swiss alpine forested sites over a combined eleven-year period were used to evaluate the role of air temperature in modeling sub-canopy incoming longwave radiation to the snow surface. Simulated sub-canopy incoming longwave radiation is traditionally partitioned into that from the sky and that from the canopy, i.e. a two-part model. Initial uncertainties in predicting longwave radiation using the two-part model resulted from vertical differences in measured air temperature. Above-canopy (35m) air temperatures were higher than those within (10m) and below (2m) canopy throughout four snow seasons (Dec-Apr), demonstrating how the forest canopy can act as a cold sink for air. Lowest model RMSE was using above-canopy air temperature. Further investigation of modeling sub-canopy longwave radiation using above-canopy air temperature showed underestimations, particularly during periods of high insolation. In order to explicitly account for canopy temperatures in modeling longwave radiation, the two-part model was improved by incorporating a measured trunk-view component and trunk temperature. Trunk temperature measurements were up to 25°C higher than locally measured air temperatures. This three-part model reduced the RMSE by up to 7.7 Wm-2 from the two-part air temperature model at all sensor positions across the 2014 snowmelt season, and performed particularly well during periods of high insolation when errors from the two-part model were up to 40 Wm-2. A parameterization predicting tree trunk temperatures using measured air temperature and incoming shortwave radiation demonstrate a simple method that can be applied to provide input to the three-part model across mid-latitude coniferous forests.