The thermal infrared (TIR) emissivity and physical temperature of snow together determine the thermal radiation of snow. The modeling of snow and ice TIR emissivity is important for climate models and remote sensing. Previous snow and ice TIR emissivity models fail in predicting the sensitivity of emissivity to snow type and snow microstructure, which was measured in experiments. Empirical models were proposed to simulate such sensitivity but not in a unified theoretical framework. In this study, we propose a snow and ice TIR emissivity model based on photon tracking by assuming that the geometric optics approximation is still valid in TIR spectral region. It is proved that the proposed model can predict both the TIR emissivity’s sensitivity to grain size for small grain sizes and the TIR emissivity’s sensitivity to snow density. These features can fully explain the experiment observed features. Moreover, the proposed model simulates snow and ice TIR emissivity in a unified theoretical framework. We also explain that the observed emissivity’s sensitivity to snow type is actually caused by the sensitivity to snow density, not grain size. This proposed model can be further used in climate models and remote sensing.