A Split-Window Algorithm for Estimating LST from Landsat-8 Satellite Images

LST and LSE are two significant parameters in climatic, hydrologic, ecological, biogeochemical, and related studies. LST is an important factor in global change studies, in estimating radiation budgets in heat balance studies and as a control index for climate models. Emissivity, is an indicator of land-cover type and resources, and also a necessary element in the calculation of LST from remotely sensed data. The main purpose of this paper is to present an operational algorithm to retrieve the Land Surface Temperature (LST) and Land Surface Emissivity (LSE) from Landsat-8 satellite images. The proposed algorithm is Split Window (SW) with band 10 (10.6 – 11.19 μm) and band 11 (11.50 – 12.51 μm) of Landsat-8 in thermal infrared range. Also for LSE mapping, the Normalized Difference Vegetation Index (NDVI) method has been suggested. This study contains two main steps: first, emissivity values of bands 10 and 11 are calculated and NDVI threshold values have been determined to separate the bare soil, fully vegetated and mixed areas from each other. Then, by using a regression relation, the values of the emissivity of the bare soil samples and mixed area have been derived. A constant value of emissivity is also used for the fully vegetated area. For a regression relation and a constant value in this study, reflectance of Landsat-8 bands has been simulated based on using two different spectral library data and relative spectral response function of Landsat-8 thermal wavelengths. ASTER Spectral Library (ASL, http://speclib.jpl.nasa.gov) and Vegetation Spectral Library (VSL), which is published by system ecology laboratory at the University of Texas at EL Paso in cooperation with the colleagues in University of Alberta (http://spectrallibrary.utep.edu/SL_browseData), were used to create simulated dataset. For validation of this step according to the lack of accurate methods for retrieving LSE from Landsat-8 imagery, the method can’t be validated with real data. Therefore, the test simulated data, which are selected randomly from simulated data, were used for validating the method. In the second step, three simulated datasets have been used. One of them for obtaining the SW coefficients and others for validating the proposed SW algorithms. The simulated datasets should include brightness temperatures, surface temperatures, emissivity and atmospheric parameters (atmospheric transmission, upwelling and downwelling radiance) for the TIRS bands. For this purpose, for each Landsat-8 TIRS band (i.e.: band10 and band11) brightness temperatures are obtained from the RTE by inversion of the Planck’s law. Surface temperatures were chosen based on the temperature of the first layer of the atmospheric profiles (T0) as T0 − 5°K, T0, T0 + 5°K, T0 + 10°K, and T0 + 20°K. The emissivity was extracted from spectral library and the atmospheric parameters have been simulated using the MODTRAN for the standard atmospheric profiles of MODTRAN (including: tropical (TRO), mid-latitude summer (MLS), mid-latitude winter (MLW), sub-arctic summer (SAS), U.S standard (USS), sub-arctic winter (SAW)). The SW algorithm coefficients for Landsat-8 were calibrated. SW algorithm coefficients were retrieved by using the least square approach based on the simulated data. Finally, this SW algorithm was tested with three datasets: simulated data, real data and satellite data. Results show that the RMSE value retrieved from the SW algorithm is equal to 1.21°k, 1.76°k and 1.03°k respectively for the three datasets. Therefore the results indicate that the proposed SW algorithm can be a suitable and robust method to retrieve the LST map from Landsat-8 satellite data.