International audience Land surface temperature (LST) is an important parameter in many research fields. Many algorithms have been developed to retrieve LST from satellite thermal infrared (TIR) measurements; of these, the most widely used are the split window (SW) and temperature–emissivity separation (TES) methods. However, the performance of the SW and TES methods can be limited by the difficulty in obtaining sufficiently accurate prior knowledge—specifically, input land surface emissivity (LSE) for the SW method and atmospheric parameters for the TES method. In this study, a procedure was proposed for selecting specific channel pairs in the TIR spectral region to accurately retrieve ground brightness temperatures without prior atmospheric knowledge, using a method similar to the SW method. Subsequently, the TES method is applied to the retrieved ground brightness temperatures to separate the LST and LSE. In numerical simulations, the three ground brightness temperatures corresponding to 8.6 μm, 9.0 μm, and 10.4 μm are acquired with an accuracy of about 0.65 K by using five channels centered at 8.6 μm, 9.0 μm, 10.4 μm, 11.3 μm, and 12.5 μm, each with a width of 0.1 μm. When inputting the three retrieved ground brightness temperatures into TES method, LST could be recovered with an accuracy of 0.87 K. Sensitivity analysis shows that LST retrieval accuracy is less affected by channel width and atmospheric downwelling radiance than by the channel center locations and channel noise. Finally, the proposed method is preliminarily applied to actual satellite data from the Atmospheric InfraRed Sounder (AIRS) and the retrieved results are compared with the pixel-aggregated Moderate Resolution Imaging Spectroradiometer (MODIS) LST product. For the study area of Australia, discrepancies between our result and the MODIS LST product appear to be about 1.6 K during the day and 1.0 K at night, indicating that the new channel configuration can be used to retrieve accurate LST from satellite measurements.