Land surface temperature (LST) is of great significance to the hydrology,ecology,environment,biogeochemical and global climate change research. Remote sensing can provide 2D land surface temperature distribution information,and can be quickly synchronously to obtain a large area of land surface temperature. Thermal remote sensing can quickly access to the LST data over a large area. Due to less atmospheric water content,LST retrieval from remote sensing data in arid areas is relatively less impacted by atmospheric condition,which enables the singlechannel algorithms achieve better inversion result. The HJ-1B satellite only acquires thermal data at a single channel, thus the comparison and assessment of the three mono-channel algorithms for LST retrieval are significant for either algorithm selection or usability test of the HJ-1B thermal data. In this study the radiative transfer equation algorithm, the mono-channel algorithm and the generalized single-channel algorithm were selected to retrieve the LST of the arid areas in Xinjiang based on the HJ-1B satellite thermal infrared images (IRS4). The results simulated by MODTRAN 4.0 show that three inversion algorithms can achieve high accuracy of LST estimation from the HJ-1B / IRS4 image data in the Shihezi and surrounding areas,with an average error of - 0.75 to 0.51 K,RMSE of - 0.17 to 0.13 K, respectively. Comparison between the LST derived from the HJ-1B data using the three algorithms and the MODIS LST product, indicates that the average error is - 1.88 - 0.83 K,and the root-mean-square error is 3.8 K with a higher determination coefficient of greater than 0.8.