Land Surface Temperature Retrieval from Landsat-8 Data: A Comparison Using a Quartz Spectral Library Based on Temperatures

摘要

Land surface emissivity (LSE) is a parameter that must be known to retrieve land surface temperature (LST). Thus, this work presents the application of NDVI Threshold Method (NDV/^{THM) in a field of coastal dunes with partial vegetation cover in Southern Brazil. We used a spectral library of quartz samples measured at different temperatures to minimize the error of the initial emissivity value in the method. Finally, we calculated LST through the improved single-channel (ISC) algorithm by using the band 10 (10,6-11,19μm) of a Landsat-8 image. The quartz emissivity was tested at two temperatures: (1) the emissivity value corresponding to the higher brightness temperature obtained from the image; (2) the emissivity value based on the saturation point of the spectral library. We verified that this approach is indispensable to acquire more reliable data. The ISC algorithm may produce an error up to 10K in the LST retrieval if the quartz emissivity is not correctly calibrated.}