The SAIL model was adapted in order to simulate spectral emissivities in the thermal infrared and compared to data extracted from the literature or from experimental works. The SAIL model was originally developed by Verhoef (1984) for simulating land surface directional spectral reflectances in the solar domain. It was adapted for simulating radiative transfers in the thermal infrared, and particularly for simulating land surface emissivity (Olioso 1995). This version of the model was called SAIL-Thermique. It required leaf area index, leaf inclination distribution function, soil and vegetation optical properties as inputs. SAIL-Thermique was evaluated against data acquired over several types of land surfaces including natural and agricultural vegetation at different levels of growth and water status. Compiled land surface emissivities for the 8-14 μm spectral band ranged between 0.92 and 0.99. Simulated 8-14 μm emissivities were favorably compared to the measurements with a root mean square difference around 0.006. When considering only herbaceous species, the root mean square difference was 0.004. In order to improve the simulation of emissivity, we updated SAILThermique on the basis of the SAIL-2M model that was developed by Weiss et al. (2001) to include the impact of the different types of vegetation organs, such as leaves, stems, ears…, on spectral reflectances. Simulations for canopies including woody plants and drying plants were performed showing that canopy emissivity can decrease significantly with the amount of woody surfaces or dry material. In a second step the model was used to assess different methodologies, including the TES algorithm and the NDVI-emissivity relationship, classically used for deriving land surface emissivity. We showed for example that the inclusion of simulated land surface emissivity spectra for surfaces including vegetation canopies had a significant impact on temperature and emissivity retrieval with the TES method. Higher emissivity estimates (by up to 0.01) and lower surface temperature (by up to 0.4 K) were obtained in the case of the ASTER sensor. Eventually, we propose a new spectral library that includes land surface emissivity spectra accounting for cavity effect as simulated by SAIL-Thermique and that can be used for training new algorithms for extracting land surface emissivity.