Relative atmospheric corrections of satellite images: multiple invariant patterns and inversions

Abstract

The correction of atmospheric effects in satellite images is a crucial task for the understanding of spatial-temporal patterns of reflectances and vegetation indices in the analysis of vegetation dynamics. In this paper the problem of inversion of radiative models of the atmosphere was analyzed by the use of invariant line patterns of soil and dense vegetation, under a scheme of multiple constraints. Radiative inversions were achieved by minimizing a merit function associated with the available restrictions and functional relationships compaction (multivariate analysis) of radiative simulations. Soil lines were better with the two constraints available in regard to the use of a single constraint. In the general case of using all available constraints for soil lines and dense vegetation, correct classification accuracies of the pair aerosol-atmosphere were greater than 80%, without prior knowledge of this information. In conclusion, when the restrictions on soil lines and dense vegetation are used to improve estimates of the multiplicative constant and optimal solutions inversion schemes are among the top five values that minimize the proposed objective functions.