Abstract: Atmospheric radiative transfer and its algorithms are the theoretical basis and effective tools in the field of remote sensing and inversion algorithm in the earth system, and also the key tools for the space,ground target recognition and quantitative assessment of background radiation. During recent decades, a series of radiative transfer(RT) model have been proposed to support a large variety of quantitative remote sensing as well as target,background discrimination research and applications. Owing to respective approximations and simplifications inherent in those RT models, their accuracy, uncertainty and adaptability are of critical significance to different researchers and end users. Validation of the RT model for its different wave band, in particular by using field experiments is necessary, especially for those applications with higher accuracy demands. Among the RT codes currently used, a considerable part of them are MODTRAN and its evolution versions. In China, MODTRAN has also been applied to the study of remote sensing, atmospheric correction of satellite images, and a wide range of applications in the atmospheric sciences, hence, the validation mainly focuses on MODTRAN model and the thermal infrared window 8—14 μm (714~1250 cm^-1) band first of all. Due to little atmospheric absorption in the infrared window band and very low radiance, this band is a range of wavelengths to which the Earth's atmosphere is relatively transparent, and is an important band used for space,ground target recognition, and ground/satellite-based remote sensing as well. Because the spectral composition of radiation transfer varies greatly with varying local environmental conditions, such as aerosol characteristics, water vapor content, surface temperature, greenhouse gases and so on, the accuracy that MODTRAN demonstrates should be attained by making the comparisons between observed radiances and the radiances computed from coincident in situ profile data. For field experiment validation, a scheme is proposed, using a special patented tethered balloon as platform and a combined sensor system consisting of both meteorological (GPS radiosonde, aerosol particle spectrometer, ozonesonde) and radiation observation instruments (visible and broadband thermal infrared imager), as well as wireless receiver,transmitter. Field experiments are conducted in August 2006 at IAP's Xianghe Observatory. During the process the tethered balloon going up and down in the atmosphere of boundary layer, measurements of both meteorological and radiation instruments at different height are carried out simultaneously. Using the observed meteorological parameters as input to RT model (MODTRAN 4.0), comparisons between observed radiances and radiances output from the model are used to validate the accuracy of the RT algorithm. The balloon is launched and drawn back for 16 times to do the validation. Analysis on the experiment results show that in thermal infrared wave band, the statistical results of the root-mean-square error of relative error between model output (with real-time meteorological parameters as input) and simultaneous radiance measurements is less than ±3%