Spatial resolution requirements for the application of Temperature and Emissivity Separation (TES) algorithm over urban areas

Abstract

Current thermal infrared satellite images are full of mixed pixels. This work is a quantitative analysis, based on radiative transfer modelling, of the distribution of mixed pixels and their impact on the use of temperature and emissivity separation (TES). TES was applied to radiance images of the cities of Basel and Heraklion simulated at different spatial resolutions by the DART radiative transfer model with 3-D representations of these cities. The accuracy of the TES was assessed by comparing the retrieved land surface temperature and surface emissivity to the input temperature and emissivity of DART. The spatial resolution of 30 m appeared to be a crucial threshold for the presence of pure pixels in these cities. When the spatial resolution reaches 30 m, the percentage of mixed pixels shows significant growth. We evaluated the performance of the TES algorithm on pure and mixed pixels. For homogeneous, isothermal, flat, and shadowless pure pixels, the variation of TES accuracy with the resolution is not obvious. For mixed pixels or pure pixels with a high nonplanar structure, the accuracy of TES even decreases with the increase of resolution. The reason may be that higher spatial resolution enhances spatial heterogeneity (due to shadow and pixel nonplanarity). A physically acceptable average temperature and average emissivity can be obtained even if TES is applied to mixed pixels. Our study stresses the need to consider the spatial resolution variation effect when applying the TES method to urban areas.