Land Surface Temperature Retrieval From Landsat-8 Data With the Generalized Split-Window Algorithm

Abstract

This paper addresses land surface temperature (LST) retrieval from Landsat-8 data using the generalized split-window (GSW) algorithm. First, radiative transfer modeling experiment is conducted using the moderate spectral resolution atmospheric transmittance algorithm and computer model fed with SeeBor V5.0 atmospheric profile database to build a data set of LST related to brightness temperatures in Thermal Infrared Sensor (TIRS) bands 10 and 11, land surface emissivities (LSEs), and total precipitable water (TPW). Then, the GSW algorithm’s coefficients are obtained through linear regression, in which the simulated data are grouped into several sub-ranges to improve the accuracy. The impacts of noise equivalent temperature difference and uncertainty of LSEs and TPW on derived LST are evaluated. Next, the TIRS channels 10 and 11 are inter-calibrated against the channels of infrared atmospheric sounding interferometer on board Metop-B. After that, LST is retrieved from the re-calibrated and clear sky Landsat-8 data using the GSW algorithm, where LSEs are estimated from the measurements of operational land imager on Landsat-8 by the modified normalized difference vegetation index (NDVI) based emissivity method, and TPW is extracted from the european centre for medium-range weather forecasts reanalysis data. Finally, the retrieved LST is cross-validated with the MOD11_L2 V6 product. The results show that the GSW algorithm can accurately retrieve LST from Landsat-8 data, and errors mainly come from the uncertainty of LSEs and TPW. Against the MOD11_L2 V6 product, the LST errors are −1.45 ± 0.80 K and −0.49 ± 0.78 K before and after the correction of LSEs and TPW, respectively.