Today I’m going to discuss another Russian gridcell 62.5N; 77.5E, which is one of the gridcells that was identified by IPCC as having a particularly strong trend . Warwick Hughes previously analyzed this cell because of this here, with this map highlighting the location of the Tarko-Sale gridcell in the top left corner of the map.
The following analysis is based on the GHCN station Tarko-Sale (23552), a town with a population of 18,500 founded in 1932, about 560 km to the south-east of Salekhard. Anorther GHCN, Aleksandrovskoe, is in the gridcell and is discussed by Warwick, but not discussed here. I may return to it on a future occasion.
We’re getting to gulag country here and the date of the establishment of Tarko-Sale seems gulag-esque. In browsing information on the area, I noticed some interesting local iore, including the use of max-min thermometers by criminals as a murder weapon. So we’d all better be careful not to rile the meteorologists.
In the early 1950s, Stalin decided to extend the Russian railway system from Salekhard to Igarsk using gulag labor. The work was abandoned, leaving little modern trace. An interesting account here describes the construction camps, mentioning in passing that:
Criminals could pour mercury from a thermometer into the ear of a sleeping man; in the morning he would simply not wake up.
Who knew that thermometers could be such a deadly weapon? A new answer to the Clue board game – Dr Jones in the library with a thermometer.
Anyway, back to statistics. My emulation of the HAdCRU3 gridcell using the Tarko-Sale GHCN v2 data is pretty close. The periods of coverage are almost identical. In the figure below, I’ve illustrated three data sources:
1. HadCRU3 gridcell 62N; 77E (annualized) 1936-2006
2. GHCN v2. This has 3 versions which cover 1937-2006. I’ve used the average of all three versions, which slightly overlap. I don’t know where the 1936 value in HadCRU comes from – maybe there’s another station in play.
3. GHCN daily data from 1995-2006 from which I calculated monthly and annual averages.
A few obvious conclusions from this emulation exercise. First, I’m able to calculate something that looks a lot like the HadCRU3 gridcell using third-party data. Second, HadCRU3 values in the 1930s and early 1940s are 0.5-1 deg C colder than the corresponding (unadjusted) GHCN v2 versions. So far I’ve not assessed whether GHCN adjusted this particular series nor why they did so, if they did. Third, and this is interesting, while the HadCRU3 gridcell values in the 1990s and 2000s are close to the GHCN v2 values, both run consistently warmer than my calculations from GSN daily data.
Finally, for what it’s worth, regardless of version, this particular gridcell does not have give me the impression of a statistically significant “trend”. There are a couple of warm episodes in the latter part (seemingly somewhat enhanced by the GHCN and CRU “adjustments”), but the graph itself looks stochastic and ends on a low note. If this your mutual fund value, you would probably not be impressed by your salesman telling you that it demonstrated consistent growth.
Here’s the picture from IPCC (2.5 vintage) showing the elevated trend in this gridcell.
Warwick Hughes said that this graphic was based on Jones 1994, for which data is available. Warwick reported here that:
The Jones 1994 data for Tarko Sale extends for only 25 years — from 1967 to 1992 — and has gaps totalling 56% of the record. It is not apparent why the thirty years of data prior to 1967 was not used. The years that have been selected result in a warming trend which would not have existed if the full length of Tarko Sale data had been used.
You can see the validity of Warwick’s comment by inspecting the above graphic. What makes the Jones exclusion in the earlier data base even more puzzling is that the earlier data was available in NDP048 (although I’ve not checked GHCN v1 yet.)