Yesterday, we had two curiosities from comparing GHCN data to HadCRU3 data – the apparent inconsistency between the HadCRU3 version in some gridcells e.g. 27.5N 117.5E and the apparent termination of much GHCN station data in 1990. Accordingly I collected all stations in the GHCN v2 data base from this gridcell and compared them with, as so often in climate science, surprising results in even the most mundane task.
In the GHCN data base, there were the following 20 stations in gridcell 27.5N 117.5E: CHANGTING, FUZHOU, GUANGCHANG, JIANYANG, JINGDEZHEN, JIUJIANG, JIUXIANSHAN, LISHUI, LONGYAN, LUSHAN, NANCHANG, NANCHENG, NANPING, PINGTAN, PUCHENG, QIXIANSHAN, QUXIAN, SHAOWU, TUNXI, YONG’AN.
Here is a re-plot of GHCN versions of the two Jones et al 1990 stations in this gridcell (Fuzhou, Nanchang) against the corresponding HadCRU3 gridcell. The HAdCRU3 values in the early 20th century are much colder than indicated by the Fuzhou station data or the few Nanchang measurements. There is also a substantial discontinuity between the periods for which the HadCRU3 series has values.
In order to see whether this discontinuity was a result of one of the other stations, I plotted all 20 GHCN versions on one graph as shown below with the HadCRU3 version marked by black points. None of the 20 GHCN series had values that corresponded to the HadCRU3 version. The only station series that went into the early 20th century was the Fuzhou station (light grey here – already plotted above and which prompted the inquiry.) Fuzhou doesn’t go quite as early as the HadCRU3 gridcell version, so there’s still a mystery here as well.
At the recent end of the data, you’ll notice that only one series (magenta) emerges from the spaghetti graph. This is Nanchang – the other Jones et al 1990 station from this gridcell. Values for 19 out of 20 GHCN stations stop in 1990. I browsed other gridcells and the same pattern is very consistent: many gridcells will have many stations, but, in China, it appears that, for many gridcells, all but one station per gridcell stop in 1990. It is amazing when you see the data.
GHCN says that many station values are updated “irregularly”. I suspect that many people would have thought that an occasion would have arisen during any of the IPCC reports since the reign of Bush I that might have occasioned an update. Buy, hey, it’s climate science, and I guess that Phil Jones has been too busy, what with IPCC and all that, to actually update little details like Chinese station data.
In deference to Michael Mann’s worry about heavy lifting, inclusion of Chinese station information after 1990 will not require the transportation of heavy equipment like tree ring borers to “remote” locations.
If the Team is too busy to update the Chinese station data, dare one ask what the CRU contract with the U.S. Department of Energy contract actually requires them to do? 95% of the Chinese station data was collected by the U.S. itself ( see the TR055 program) nearly 20 years ago. What has the CRU been doing about Chinese data since then? Were they so convinced by Jones et al 1990 about Chinese UHI that they felt entitled to stop collecting information other than one big Chinese city for most gridcells? The Team never ceases to amaze.
There’s an obvious moral to this story. Ross McKitrick observed a long time ago that the Consumer Price Index is never be calculated by a bunch of academics at a university. It’s calculated by a proper statistical service who don’t mind doing little things like actually updating 95% of the Chinese data more frequently than once every 20 years. So aside from being surly and unresponsive with horrendous documentation of what they did, CRU doesn’t seem to be very efficient either. I wonder what justification the U.S. Department of Energy has for giving this inefficient program another cent.
And, not to be too distracted by CRU (and GHCN inefficiency), what of the adjustment that prompted the inquiry in the first place? There’s only one station in this gridcell that goes into the early 20th century Fuzhou – and its early 20th century values are relatively high. What is the explanation for the low values in the HadCRU3 gridcell? In other gridcells, I’ve been able to tie the HadCRU3 gridcell value to station values, but not here. Here we run into the CRU cone of silence: you have to quasi-litigate to try to get information on what stations were used in the gridcell; then you’d have to litigate to find out what they did and probably never really succeed. And by then, they’d have “moved on” and substituted some new and equally undocumented result.
Update: here’s the NOAA gridded value for this gridcell compared with HAdCRU3. The Fuzhou values in the early part as well as the Nanchang values in the later part can be discerned in the NOAA gridcell. What did CRU do?