Addressing information content limitations in the inversion of remote sensing observations: strategies of evaluation and optimizing aerosol models using GRASP algorithm

Abstract:

One of the main challenges in the development of remote sensing retrieval is adequate addressing of limitations in the observation information content. Indeed, the information content with the respect to atmospheric parameters changes strongly depending on the type of measurements. For example, spectral, multi-angular, polarimetric, satellite or ground-based, passive or active measurements have very different sensitivities to aerosol properties. Correspondingly, the number and type of retrieved characteristics need to be adequately chosen for each specific observation.

GRASP (Generalized Retrieval of Aerosol and Surface Properties) is a versatile algorithm developed by Dubovik et al., [2011, 2014] for deriving atmospheric parameters from diverse observations. Several useful strategies have been developed in frame of GRASP for optimizing forward model and the parameter set to be retrieved. First, the amount and type of a priori information used by GRASP can be changed depending on the application. For example, (i) for any retrieved parameter direct a priori estimate can be used, (ii) to limit variability of retrieved continues functions, such as size distribution, spectral dependence of refractive index, etc., a priori smoothness constraints can be applied and (iii) once a group of coordinated observations is inverted simultaneously, e.g. satellite pixels, multi-pixel a priori smoothness constraints can be used for liming spatial and temporal variability of parameters retrieved in different pixels. Second, many aspects in GRASP forward model can be changed depending on the application. For example, parametrization of different complexity can be used for characterizing aerosol particle size and shape distributions, composition and vertical profiles. Third, GRASP allows rather straightforward evaluation and testing of the developed retrieval by applying it to the observations with higher information content. For instance, the simplified aerosol models employed in satellite retrieval can be adapted for processing AERONET ground-based observations alone or combined with satellite data. Such approach allows identifying shortcomings in the assumptions and finding fruitful alternatives. The use of the concepts is illustrated by applications to satellite and ground-based observations.