Lacking that, I would say that in order to compute temperature change as a function of changes in CO2, solar irradiance, or other forcings, we do need to know absolute temperatures, but at the level of each grid space in a model analysis rather than as a global mean.