Land surface temperature retrieval from thermal infrared data: An assessment in the context of the Surface Processes and Ecosystem Changes Through Response Analysis (SPECTRA) mission

[1] SPECTRA (Surface Processes and Ecosystem Changes Through Response Analysis) is one of the core candidate missions which is being proposed for implementation in the European Space Agency (ESA) Earth Explorer program of research oriented missions. The scientific objective of the SPECTRA mission is to describe, understand, and model the role of terrestrial vegetation in the global carbon cycle and its response to climate variability under the increasing pressure of human activity. The SPECTRA satellite will embark an optical hyperspectral payload covering the solar spectral range (0.4 to 2.4 mm) and thermal infrared region (10.3 to 12.3 mm). This paper is focused on the land surface temperature retrieval from SPECTRA thermal infrared data. In the first part of the paper, generalized single-channel and split-window methods are discussed and compared, showing that single-channel methods provide similar or better results than split-window methods for low atmospheric water vapor content, whereas split-window methods always provide better results for high atmospheric water vapor content. In the second part of the paper, split-window and dual-angle algorithms have been developed for SPECTRA thermal channels. A sensitivity analysis of the algorithms has been also carried out, revealing total errors for split-window algorithms of around 1.5 K. For dual-angle algorithms, total errors less than 1 K are obtained when the combination nadir-60 is considered. Finally, a dual-angle algorithm for sea surface temperature retrieval has been developed for different view angles. The study of the variation of the total error with observation angle allows estimation of the best nadir-forward combination. Hence an optimal forward view of 52 referred to the observer zenithal angle (or 45 for satellite view angle) has been obtained, leading to an error of 0.4 K when the sensor noise error is 0.1 K and 0.3 K when the sensor noise error is 0.05 K.