Because it's always a model, and has plenty of parameters to tweak, there is no "measuring the temperature of the earth at all points at the same time."
Yeah, that makes sense—I thought op was suggesting it wasn’t primarily from observed data. Still, it seems clear that it should be possible to make conclusions about average temperature and trends from a large set of collected data, if it the statistical methodology is rigorous. Simply casting doubt (as it seems like the op is doing) without offering anything to question those methodologies adds nothing of value to the discussion.
