Land Surface Temperature Retrieval and Result Analysis Based on Landsat Imagery in Dianchi Basin

Surface temperature is an important indicator for the global environment analysis, and remote sensing technology has been widely used for surface temperature retrieval. At present, the main algorithms for surface temperature retrieval include: Radiance Transfer Equation (RTE), Mono-window Algorithm (MWA), Subpixel Decomposition Techniques. However, these algorithms' comparative analyses are insufficient. In this paper, according to the above limitation, the Dianchi Lake Basin as study area and Landsat-5/7/8 satellite remote sensing images as the main date sources, the three algorithms are used to complete surface temperature retrieval and analyze the differences. Finally, the retrieved results' accuracies of different algorithms were verified using land surface temperature of daytime (LTD) data. The results showed that: (1) The correlation coefficient of the three algorithms retrieved results with LTD were both higher than 0.75. The similarity coefficient between the retrieved land surface temperature results of three algorithms were greater than 0.98, both of them could represent the spatial distribution of surface temperature and the differences of different underlying surfaces. (2) The correlation between the results of the radiance transfer equation method and LTD reached 0.804 (Landsat-8 OLI/TIRS), which was suitable for high-precision surface temperature analysis. (3) The mono-window method used atmospheric transmittance and atmospheric average temperature to participate in the calculation, which could retrieve the surface temperature quickly and effectively. (4) The emissivity modulation (EM) method could retrieve the surface temperature effectively by adjusting the emissivity, and the correlation coefficient between its results with the LTD data reached 0.8003 for Landsat-8 OLI / TIRS image. The emissivity modulation method only needs to calculate the surface emissivity before it can retrieve the surface temperature with known parameters. It's suitable for the surface temperature retrieval using remote sensing data before 2000 years and the automatic surface temperature extraction of large-scale remote sensing data.