earth-window-mismatch-is-an-alignment-problem

# Earth: a window mismatch is an alignment problem, not a reason to decline A comparator that does not line up with your result - different period, different baseline epoch, different initial state, different unit - is the normal case in this field, and aligning it is part of the analysis rather than a precondition for it. Every time you are about to write "no comparable estimate exists for our window", check that you have not just described an offset. ## The failure this prevents A run was handed several model-projection papers alongside its observational archive. It read them properly: it converted their ensemble aggregates into its own unit with the constant those papers cite, found that their published horizon begins years after its own record starts and is stated relative to an initial state at that later date, and quoted the sentence saying the models' pre-horizon trajectories are mutually inconsistent. It then closed the whole comparison in a design note - "no ensemble aggregate exists for the observed window even in principle" - and substituted a scalar agreement test against a single historical calibration target, on the argument that the models calibrate to it. Because the clause was recorded as *declined with evidence*, no later stage reopened it. The shipped report contains no projection curve, no pathway, and no horizon past the last observed year. The source study did the comparison anyway, and said how in one clause of a caption: the projections were **offs...