So Eli was slumming over at Lucia's where Steve Mosher was doing his best tipsy Richard Tol imitation and the subject was the Cruz Pause
18 year no trend in the satellite record, when a thought occurred. (Ear tip to Tamino, also look at his follow on, Drift)
If the changes in temperature over short periods (like days or months or even annually) track each other, even just in direction in the satellite and surface records (so) then that is pretty convincing evidence that the problem is a long term drift in one or the other and that on the short term they are measuring the same thing.
So Eli hit Wood for Trees and compared the RSS land only record with CRUTEM4 between 2005 and 2015. (RSS offset by 0.25 K)
If there was significant random (not actual) variation in either record, one would expect that there would be many months when the two curves moved in opposite directions. There are a few, but it's a lot like finding the panda. From this we conclude that MSU and CRUTEM4 are consistent on a monthly and even an annual basis. It might be even more interesting to look at this on a daily basis and even match times and AMSU footprint areas
Thus if (see Nick Stokes) the long term trends diverge, and the short term anomalies agree, that is pretty good evidence of systematic drift.
The surface record analysis is simpler and a drift would require correlation between drifts at many stations using different instruments that are calibrated on site.
Eli concludes that the drift is most likely in the AMSU satellites or the processing of the AMSU data. Unanticipated aging of the receiver or the internal hot calibration target seems to Eli most likely, although there might be something involving orbital decay (less likely now because this caused a lot of trouble early on) or even changing land/sea/ice patterns which affect the AMSU response.
Since there are four or five satellites carrying the AMSU units and there is a newer ATMS system, analysis of where the discrepancy enters would not be simple.