Ermida, SL; Trigo, IF; Holley, G; DaCamara, CC (2020). A multi-sensor approach to retrieve emissivity angular dependence over desert regions. REMOTE SENSING OF ENVIRONMENT, 237, 111559.

Land surface emissivity is a directional quantity and most natural surfaces are anisotropic emitters. Experimental studies show that for homogeneous surfaces like bare soils, the emissivity decreases with view angle. Translation of the directionality of emissivity as obtained from experiments or models to satellite spatial scales of kilometers or more is challenging given the natural heterogeneity of the land surface at those scales. We propose a multi-sensor approach to retrieve TIR emissivity angular dependence from satellite observations. The method is based on simultaneous radiance measurements over stable and homogeneous targets, performed by multiple remote sensing platforms in low-earth orbit (MODIS/VIIRS) and geostationary orbits (SEVIRI) with different viewing configurations. In the case of homogeneous scenes, we assume that differences in observed brightness temperatures are exclusively due to differences in the atmospheric optical path and in surface emissivity within each view angle, as surface temperature is the same for both sensors. We then use radiative transfer simulations to derive the emissivity changes between the sensors due to the viewing position. It is shown that emissivity values decrease with view angle for all thermal infrared channels considered, with the highest variations for channel centered at 8.7 mu m, followed by 10.8 and 12.0 mu m, respectively. The MODIS MYD21 product agrees with the multi-sensor approach for the 8.7 mu m channel but shows negligible dependence on view angle for 10.8 and 12.0 mu m bands. Results suggest that the emissivity calibration curve used in the MYD21 algorithm restricts the retrieval of the full angular variation of emissivity at these channels. A sensitivity analysis shows that the proposed multi-sensor method is robust, with reduced impact of the uncertainty of the atmospheric correction on the retrieval. The impact of emissivity angular dependence on retrieved LST may be as high as 3.0 and 1.0 K, respectively, for the GSW and the TES algorithms for angles up to 65 degrees.