Thermal band selection for the PRISM instrument: 1. Analysis of emissivity‐temperature separation algorithms

One of the missions being planned by the European Space Agency (ESA) within the framework of its Earth Observation Programme is the Processes Research by Imaging Space Mission (PRISM). The PRISM instrument consists of a thermal sensor whose main objective is to retrieve accurate land surface temperatures (LST) and whose band positions are 3.5–4.1 μm, 8.1–9.5 μm, 10.3–11.3 μm, and 11.5–12.5 μm. We have studied the optimal design of this instrument to retrieve accurate LSTs. First, we have analyzed several emissivity‐temperature separation methods (part 1) and atmospheric and emissivity correction algorithms (part 2). Finally, we have identified the optimal band configuration (part 3). This paper is the first of a series of three and addresses the question of the emissivity‐temperature separability. Among all the existing algorithms, we have studied the “absolute methods,” which are able to estimate the absolute value of emissivity at satellite scale and can yield better results in the emissivity estimate. These methods are the algorithm based on the temperature‐independent thermal infrared spectral index (TISI), the alpha coefficients method, and the algorithms which use visible, near‐infrared, and shortwave infrared data to estimate the thermal emissivity (vegetation cover method (VCM)). The study consisted of an analysis of both random and systematic errors of each method. The results indicate that emissivity can be obtained with an error of ±1.7–5% using the alpha coefficients, ±1.7–3% using the TISIs, and ±0.5–1.4% using the VCM, depending on the spectral region. In all cases the error decreases with wavelength, and the lowest errors are achieved in the 10–12 μm spectral region, due to small variability of emissivity. It is necessary that the two first methods use radiosondes simultaneous with the satellite overpass to perform the atmospheric corrections on the thermal data; in addition, they show important sources of systematic errors, which will increase the uncertainty in the emissivity estimate (even in the best possible case). The VCM does not use radiosondes and does not present important sources of systematic error. It appears to be the procedure with the most favorable error propagation characteristics. Thus the VCM could be the most adequate method for retrieving the land surface emissivity (LSE), within the framework of this work.