Land surface temperature retrieval from Landsat 8 thermal infrared data using mono-window algorithm

Abstract

Land Surface Temperature (LST) is a significant surface biophysical variable. This parameter is also important in various fields such as urban thermal environment, agricultural monitoring, surface radiation, and energy balance. Data from Landsat satellites are vital remote sensing data for LST retrieval since the 1980s. The present Landsat 8 Thermal Infrared Sensor (TIRS) imagery provides a new data source for LST retrieval. Landsat 8 TIRS is improved compared with the Landsat 6 Thematic Mapper. Landsat 8 data are extensively applied, so the mono-window algorithm should be updated with new sensor characteristics. Therefore, we aim to explore an adaptive method with more reliable accuracy to retrieve LST using Landsat 8 TIRS data. In this paper, a relation model (TIRS10_SC)was established between LST and several parameters, namely, brightness temperature, mean atmospheric temperature, atmospheric transmittance, and land surface emissivity. The model was based on the radiative transfer equation and characteristics of Landsat 8 TIRS10.The LSTs of the study area were retrieved by initially deriving the atmospheric transmittance from MODIS data and MODTRAN simulation results. Then, the mean atmospheric temperature was obtained using empirical formulas, and land surface emissivity was retrieved from the Landsat 8 OLI data using image classification-based method. Finally, the LSTs of the study area were retrieved from the processed data. The algorithm and retrieval results were assessed by simulated and measured data. Meanwhile, the sensitivity of variables in themono-window algorithm was analyzed. Results show that the mono-window algorithm can perform well for Landsat8 TIRS data for LST retrieval. The LSTs of different land-cover types in study area varied. The LSTs of bare soil and cements were evidently higher than those of the vegetated areas. The LST of the former varied between 24.12 ℃ and 32.25 ℃, whereas that of latter ranged from 10.72 ℃ to 19.79 ℃. Furthermore, compared with the measured data, the average error and correlation coefficient of retrieved LSTs were 0.83 ℃ and 0.805, respectively. The accuracy of the algorithm was also assessed using simulated data, which showed that the error in the LST data in the majority of cases ranged between 0.2 ℃ and 0.3 ℃. The retrieval results agree with the assessed temperature data. Results from the analysis of the sensitivities of land surface emissivity, atmospheric water vapor content, and average temperature showed that the TIRS10_SC algorithm can obtain more reliable results with higher sensitivities for the former two para meters and lower sensitivity for the latter one. The proposed algorithm can be applied for the fast retrieval of LST using Landsat 8 TIRS data.