NASA publishes what it calls an "estimate" of the Earth's global mean surface temperature. Digging through papers I've been able to find some documentation of how these numbers are computed, but I have had a frustratingly hard time answering what seems to me a much more basic question: what is this "estimate" an estimate of? That is, what is the abstract conceptual quantity that we'd really like to know, if we could? Without that, I don't understand how we can evaluate how well, for instance, the methods they use to blend together data sources of different resolutions and reliabilities, since the definition wouldn't have any stronger "correctness" criterion than just "this is how we put these numbers together".
It seems to me that the conceptual quantity they're trying to measure is, for a given year:
- Imagine gridding up the earth into equal-area grids and putting a thermometer in each one.
- For each thermometer, imagine taking the limit of the average reading as number of readings increases for a one-year timespan from midnight on Jan 1 to just before midnight on Dec 31 of the given year
- Now take the limit of the average of all those single-thermometer limits for every thermometer in your grid, as the size of the grid squares decreases.
But this is really just a guess as to what they're after -- I have skimmed over several papers on this subject and I haven't been able to find any discussion of this at all. Is this just something basic that climate researchers know that I'm missing?