Aerosol retrieval from space – how does geometry of acquisition impact our ability to characterize aerosol properties

Abstract

For decades now, the retrieval of aerosol property has been successfully achieved from space-borne sensors from which measurements it is possible to derive specific parameters such as optical thickness, absorption, type, refractive index, or size distribution. In the reflective spectral domain, remote sensing of aerosol properties relies on the top-of-atmosphere measurement of the sun irradiance scattered by aerosols in different directions. This measure is intrinsically linked to the aerosol phase function. Ground-based measurements are made for many viewing directions providing a good description of this phase function. For satellite remote sensing, the phase function cannot be measured in so much detail. Only a single scattering angle for mono-viewing sensors, or a limited range for multi-view sensors, is accessible. The associated geometry varies very significantly along the swath, from East to West, and along the orbit, from North, to Tropics, and South. Whatever the considered aerosol retrieval approach, the performance cannot be the same from these very different geometrical configurations, and may significantly differ. This aspect is in general not well documented. In this paper, the scattering angle range distribution (ScARD) is described in the case of the EPS-SG/3MI multi-view polarimeter. Based on reference aerosol phase functions, it is anticipated how the retrieval performance could be impacted. Other cases are simulated trying to extrapolate the conclusion to other types of sensors having more limited swath or number of views. The ScARD is described for these different situations, including the variation along the orbit and along the swath. Important recommendations are drawn including the need to document the geometrical part of the information content provided by the sensor (not only spectral), to better consider the associated classes of viewing geometry for the development of retrieval algorithms (which could limit the ability to retrieve some parameters), but also for the validation of products.