Land-surface temperature retrieval from Landsat 8 single-channel thermal infrared data in combination with NCEP reanalysis data and ASTER GED product

Abstract

ABSTRACTLand-surface temperature (LST) is an important parameter in the climatological, hydrological, ecological, and meteorological studies. The Thermal Infrared Sensor (TIRS) on board the Landsat 8 is a key instrument to collect thermal infrared (TIR) data. The Landsat series sensors provide continuously acquired collection of space-based TIR data. In this study, we proposed a method for retrieving LST from Landsat 8 TIRS single-channel data. The National Centers for Environmental Prediction reanalysis data in conjunction with the Moderate Resolution Transmittance Code 5 were used to correct atmospheric effects. The ASTER Global Emissivity Database product was used to correct the effects of surface emissivity. In situ LST measurements were collected by eight and four SI-111 infrared radiometers in the study areas A and B, respectively. The in situ LST was used to validate the retrieved LST. For the study area A (sands), the bias varies from approximately −1.3 to 1.7 K, and the root mean square error (RMSE) from approximately 1.2 to 2.1 K. For the study area B (grasslands/snow), the bias ranges from approximately −1.0 to 0.4 K, and the RMSE from approximately 1.1 to 1.5 K. To further compare the retrieved LST and the in situ LST at coarser pixel scale, all of the retrieved LST and the in situ LST were, respectively, averaged as the corresponding LST at 1 km pixel scale (e.g. Moderate Resolution Imaging Spectroradiometer). The biases of the differences between the two averaged LST at 1 km pixel scale for all TIRS scenes are approximately −0.2 and −0.5 K for the study areas A and B, respectively, and the RMSE values are approximately 1.2 and 1.0 K for the study area A and B, respectively. These results indicate that the proposed method can be used to retrieve LST from single-channel TIR data with a reasonable accuracy.