On the determination of the atmospheric air light as a normally underestimated perturbation signal