Aiming at the inconsistent accuracy of Land Surface Temperature(LST) inversion from Landsat-8 imagery using Single-Channel (SC) Algorithm and Split-Window (SW) Algorithm in different regions, the inversion accuracy of the Generalized Single-Channel (GSC) and SW algorithms proposed by Jiménez-Mu?oz et al. were compared and analyzed based on the in-situ measurements of three NOAA’s Surface Radiation Budget Network (SURFRAD) sites with different land cover types and altitudes, and 37 high quality Landsat-8 imagery from January 1, 2018 to June 30, 2019. The results show that there is a certain seasonal dependence of the inversion accuracy. In spring and winter, the Mean Absolute Error (MAE) of GSC algorithm is about 1 K, and the value of SW algorithm is higher than 1 K. The Root Mean Square Error (RMSE) of GSC algorithm is about 0.37 K and 0.13 K lower than that of SW algorithm, respectively. The accuracy of GSC algorithm is about 0.53 K higher than that of SW algorithm in spring, and about 0.14 K higher in winter. In summer and autumn, the MAE of both algorithms is less than 2 K, but SW algorithm has higher accuracy. Among them, the accuracy of GSC algorithm is about 0.35 K lower than that of SW algorithm in summer, and about 0.25 K lower in autumn. The RMSE of GSC algorithm is about 0.24 K higher than that of SW algorithm in autumn. The correlation coefficients between the two algorithms and the measured LST are 0.9907 and 0.9922 respectively, showing relatively strong correlations, and SW algorithm outperforms GSC.