Earlier today Steve McIntyre posted a paper drawing attention to a discrepancy in the CRU’s treatment of Siberian temperature series may have a tendency to overstate the warming trend.
It turns out this was one of those papers that Phil Jones went to town on to keep critical papers out of the published literature. In that paper it mentioned the rural station of Kirensk as one that should be looked at because it has a long record and isn’t subject to urban heat island effects. This is interesting because this station is one that the CRU released as part of their subset of recently released data and it is available in the GHCN records.
So I downloaded the data for both and compared them:
As you can see the GHCN and CRU versions agree fairly closely from about 1950 but CRU has a noticeable warming trend because it is significantly colder in the past. How they justify this sort of adjustment is interesting but unknown since they don’t release their methods. I assume this is some sort of homogeneity adjustment that aligns the trend of this station with neighbours (which may be increasing due to urban effects).
Also note that GHCN also has an “adjusted” series for this station, but those adjustments are minor – usually less than 1/10 of a degree.
That graph provides no less than three temperature series for one station, but it doesn’t tell us which one (if any) are actually right. There is actually one more series for that station, and that is the GHCN Daily measurements. The other series are already monthly averages, and have been processed somewhat. The daily observations, on the other hand, are supposedly completely raw, recording exactly what Ivan wrote down on his form as he trudged out into the Siberian winter to read the thermometer.
Unfortunately, that data is missing the records for the early part of the century, but plotting it is interesting nonetheless:
The station reliably reported from about 1935 to about 1999, after which it has a large number of missing values, which would explain the larger variance from the year 2000 on. Presumably CRU and GHCN estimate the missing days from neighbouring stations and calculate the monthly values from a mix of real observations and estimates.
But both CRU and GHCN show smaller variances from 1935 to 1999, too, even though there are essentially no missing values to estimate. So it appears that even the GHCN “raw” data isn’t really raw, but has been subjected to a degree of homogenization.
So what is the “real” temperature trend of Kirensk and, by extension, southern Siberia? It looks like it can be whatever you want it to be.
