Interpretations of an adjoint-derived observational impact measure