This could easily be a business case: Two managers. One sits in his office, looking at spreadsheets, trying to figure out if the factory is doing OK. The other spends most of his time on the factory floor, trying to see what is going on. Both approaches have value, and both have shortcomings.
Shift the scene now to the physical sciences: Two geologists. One sits at his computer looking at measurement data sets, trying to see trends through regression, interpolation, and sometimes via manual adjustments and corrections. The other is out in the field, looking at physical evidence. Both are trying to figure out sea level changes in the Maldives. The local geologist can’t see global patterns, and may have a tendency to extrapolate too broadly from a local finding. The computer guy doesn’t know how his measurements may be lying to him, and tends to trust his computer output over physical evidence.
It strikes me that there would be incredible power from merging these two perspectives, but I sure don’t see much movement in this direction in climate. Anthony Watts has been doing something similar with temperature measurement stations, trying to bring real physical evidence to improve computer modellers correction algorithms, but there is very little demand among the computer guys for this help. We’ve reached an incredible level of statistical hubris, that somehow we can manipulate tiny signals from noisy and biased data without any knowledge of the physical realities on the ground (“bias” used here in its scientific, not its political/cultural meaning)