For details of what is illustrated here, start from Wanderings of a Marseille January 1978 temperature, according to GHCN-M, continue with GHCN-M: Stations similar to Marseille/Marignane Pairwise Homogenization may perform better in Canada and the United States of America. This is not unexpected − pairwise homogenization was benchmarked for the United States. (C. N. Williams, M. J. Menne, P. W. Thorne, Benchmarking the performance of pairwise homogenization of surface temperatures in the United States. J. Geophys. Res. 117 (D5), D05116 (2012). doi:10.1029/2011JD016761). Further discussion to come later once images have been generated and added for remaining regions.
Update: As I have started to archive USHCN data as well as GHCN-M data, an example of USHCN adjustment is also shown below for one station (Sacerton).
403 Canada (847)
Starting with the station which shows the greatest difference overall between unadjusted and adjusted values, although this one does not resemble Marseille/Marignane. The adjusted values in fact appear to show the pattern of small variations until 2014 which might be expected of a properly performing homogenization process, then suddenly adjustment ceases and the adjusted values revert to the unadjusted value. Why? “Nitchequon” in the GHCN-M data is not in fact Nitchequon, but rather Nitchequon until 1985, with later data added from a station much further south, Pangnirtung. Until late 2013 GHCN-M added data from Pangnirtung from 1996 onwards. Then for a time Nitchequon reverted to being purely Nitchequon. Then sometime around March 2015 data for 2015, sometimes quality flagged, sometimes not, reappeared. As I wrote in a blog post A few more GHCN problems back in 2012, what has happened is that Canada has recycled some WMO id numbers. 71826 was previously Nitchequon but is now Pangnirtung. The impressive 5°C cooling adjustment imposed on the real Nitchequon data is needed to bring it into line with the data from Pangnirtung, much further south (there is a map in my 2012 post).
There are a number of other Canadian stations in the GHCN-M inventory which have changed WMO identity in similar fashion, but (fortunately) no data from the new WMO identity has (as yet?) been added to the data records, which ended decades ago.
405 Costa Rica (4)
406 Cuba (14)
407 Dominican Republic (28)
408 El Salvador (4)
410 Guatemala (6)
412 Honduras (12)
413 Jamaica (5)
414 Mexico (119)
415 Nicaragua (7)
417 Saint Kitts And Nevis (1)
423 The Bahamas (3)
424 Trinidad And Tobago (3)
425 United States Of America (1921)
As US stations make up more than 25% of the GHCN-M inventory, it seems appropriate to include rather more US stations here than are included for other countries. With the large number of stations available for inclusion, I have decided that a better criterion for inclusion would be volatility of adjustments under the current GHCN-M version, v3.3.0
For other countries examined so far I have looked more at differences from one version to another, and adjustments which switched frequently above and below the unadjusted value. I may return later to countries already examined and consider adding stations which show considerable volatility under the current version v3.3.0
Many of the stations shown above are also USHCN stations. I have not up to now paid much attention to USHCN, as the US is only a small part of the global land mass. USHCN however uses the same Pairwise Homogenization as GHCN-M, and this is regarded as performing more reliably with the greater station density and station quality of the USHCN network. I have therefore now started to archive daily USHCN datasets as well as GHCN-M datasets, and will include the USHCN adjusted values in future images for USHCN stations some time next month (September), once I have archived sufficient USHCN datasets. USHCN adjusted values for USHCN stations can differ from GHCN-M adjusted values for the same stations as the GHCN-M network includes other US stations outside the USHCN network as well as those included in the USHCN network. With the latest USHCN and GHCN-M versions the adjusted values seem to be the same when both USHCN and GHCM-M have adjusted values, and differ only when only one adjusted value is calculated. More on this later.
For now, here is Sacaton, the first US station illustrated above, with USHCN values added. The data used continues here to August 31st 2015. GHCN-M values are missing for August 29th, although USHCN values are shown. For some reason my scheduled archiving job failed to download GHCN data for that day.
427 Bermuda (U.K.) (2)
429 Cayman Islands (U.K.) (1)
431 Greenland (Denmark) (15)
432 Guadeloupe (France) (1)
433 Martinique (France) (2)
434 Netherlands Antilles (Netherlands) (2)
435 Puerto Rico (U.S.A.) (17)