Abstract The current scientific understanding of top-of-atmosphere globally-averaged albedo stability and hemispheric symmetry on interannual to centennial time-scales is largely phenomenological and diagnostic, leading to diverging albedo estimates over the 21 s t Century. We present a framework for hypothesis testing and theory development, complemented by observations, data methods, and models, to advance towards a theory for these phenomena. We present the theory’s necessary features needed and then explore two hypotheses and their observational tests: that albedo is (1) constrained by invariant Earth system properties, or (2) maintained by cloud buffering. We argue that there is a supporting role for artificial intelligence (AI) methods that express the non-linear processes that underlie these observed phenomena. Additionally, we show that the response to and relaxation from major perturbations like stratovolcanic eruptions may help to falsify hypotheses. Finally multi-model ensemble experiments and observations can support this multi-pronged approach to advance towards a working theory for albedo.
Feldman et al. (Tue,) studied this question.