Researching Of Remote Sensing Inversion Method On Land Surface Emissivity

The land surface emissivity (LSE) is one of important parameters in temperature inversion from thermal infrared remote sensing, which is popularly applied to the process of water and heat exchange and radiative transfer. This paper used TM and MODIS data as the data source for research, a comparative study has been carried out on the four algorithms to estimate the land surface emissivity based on the high linear relationship between the Normalized Differential Vegetation Index (NDVI) value and LSE, we also summarized the basic characteristic of all kinds of LSE and examined sensitivity analysis of retrieving land surface emissivity (LSE) on the basis of NDVI threshold method, which was theoretically established on the vegetation index mixture model proposed by Valor and Caselles in1996, meanwhile, with the aid of MOD11C3of Terra-MODIS L3level products, we obtained10-year spatial-temporal data sets of LSE in China from2001to2010. The results show that:(1) In all literatures, the value of LSE about same features or land types were different from each other, but present concentration trend, the value of water mainly between0.9860and0.9950, the town/building between0.9609and0.9760, the bare soil radiation between0.9600and0.9750, and the vegetation was concentrated in the range of0.9800to0.9900.(2) A comparative study has been carried out on the four published algorithms (Van and Owe,1993; Valor and Caselles,1996; Sobrino et al,2003; Qin Zhihao et al,2004) to estimate the land surface emissivity, respectively analyzing the applicability and the accuracy from the pixel scale and classification scale by comparing the results of LSE and MODIS LSE products, which shows that the methods proposed by Valor and Caselles was better than the others.(3) The approach of controlling variables was used to identify the key parameters shaping significant effects on LSE retrieval from atmospheric factors, topographic factors, NDVI threshold and the emissivity value of pure pixels. The research has been discovered that the degree sequence of sensitivity:the emissivity value of pure pixels> the NDVI threshold of pure pixels> topographic factors> atmospheric factors, meanwhile, the emissivity value of pure pixels and the NDVI threshold of pure pixels has been identified as the key parameters.(4) The10-year spatial and temporal data sets of China’s land surface emissivity from2001 to2010show that Northwest China’s desert region has the minimum land surface emissivity in China, with little seasonal variations. In contrast, there are significant seasonal variations in land surface emissivity in Northeast and Northwest China, Qinghai-Tibet Plateau, the Yangtze River Basin, East and South China. In winter, the land surface emissivity is relatively high in Northwest China. The land surface emissivity of Qinghai-Tibet Plateau maintains a lower value from November to March, while it is higher in other months. The land surface emissivity of the Yangtze River Basin, East China, South China and Sichuan Basin decreases from July to October, and peaks in August. There is clear relationship between the spatial-temporal distribution of Chinese land surface emissivity and temperature:the higher the emissivity, the lower the temperature, and vice versa.