A practical algorithm to inverse land surface component temperatures from ATSR-2 and ASTER data

In this study, a thermal model-based algorithm has been developed. This linearized algorithm can invert land surface component temperatures in an ATSR pixel. The proportion of each component in an ATSR pixel is gotten using matched ASTER data. Then, assumed that the radiation of an ATSR pixel is the sum of the radiation from several components, we get a simple linear thermal model. Atmospheric effects on ATSR thermal data are removed using a split-window algorithm. Before inversion, the sensitivity of parameters is analyzed using Uncertainty and Sensitivity Matrix (USM). During the inversion process, a Multi-stage Sample-direction Dependent Target-decisions (MSDT) strategy is taken, that is, the most sensitive and uncertain parameters are inverted first by fixing some less sensitive parameters at their prior values. Compared with the general direct inversion method, MSDT strategy can get a more robust result. With other prior knowledge, this algorithm can invert soil and vegetation temperature