In his comment to How much Estimation is too much Estimation?, Anthony Watts suggested I create a scatter plot showing station distribution with latitude/longitude. It turned out not to be the ordeal I thought it might be, so I have posted some of the results in this thread. I started with 1885 and created a plot every 20 years, ending in 2005. I deliberately ended with 2005 because this is the final year in the GHCN record prior to the US station die-off of 2006.
Every dot on a plot represents a station, not a scribal record. Stations may be comprised of multiple records. A blue dot represents a station with an annual average that was fully calculated from existing monthly averages. A red dot represents a station that had missing monthly averages for that year, so the annual average had to be estimated. Stations that had insufficient data to estimate an annual average are not shown.
In the case where multiple scribal records exist for a station in the given year, I assigned a blue dot if all records were fully calculated from existing averages, a red dot if at least one record was estimated, and no dot if none of the records could produce an estimate. I believe this errs in the direction of assigning more blue dots than is deserved. Hansen’s bias method mathematically forces estimation to occur during the period of scribal record overlap.
The first plot shows coverage in 1885, five years into the GHCN record.
1905 shows improved coverage across the continental US, Japan and parts of Australia. A few stations have appeared in Africa.
1925 shows increased density in the western US, southern Canada, and the coast of Australia.
At the end of WWII, not a lot of change is noticeable other than improved coverage in Africa and South America as well as central China and Siberia.
In 1965 we see considerable increases inChina, parts of Europe, Turkey, Africa and South America.
A decline in quality seems to be apparent in 1985, as many more stations show as red, indicating their averages are estimated due to missing monthly data.
A huge drop in stations is visible in the 2005 plot, notably Australia, China, and Canada. 2005 was the warmest year in over a century. Not surprising, as the Earth hadn’t seen station coverage like that in over a century.
The final plot illustrates the world-wide station coverage used to tell us “2006 Was Earth’s Fifth Warmest Year“.
Update (Steve Mc): USHCN station information gets added into GHCN with a lag of almost a year (noted in comments below). Jerry Brennan, who’s followed this for some time, reports the following update schedule in the past:
USHCN station data for the year 2002 were published in the USHCN website by May 2003, and added to GHCN between November 8, and December 10, 2003.
USHCN station data for the year 2003 were added to GHCN between April 10, and May 6, 2004, and published in the USHCN website by January 2005.
USHCN station data for the years 2004, 2005, and the first three months of 2006, were added to GHCN between August 13, and September 11, 2006, and published in the USHCN website, with data through October 2006 by March 2007. The additional months of data were not added to GHCN.
USHCN station data through May 2007 were published in the USHCN website in October 2007, but the “new” data have not been added to GHCN as of this date.
By the end of February of each year, GHCN will usually have data for the full previous year from only 120 (non USHCN) stations in the 48 contiguous USA states.
I might add that there are two locations for USHCN data, one at NOAA and one at CDIAC. The NOAA version is more updated than the CDIAC version – perhaps there are other differences as well. I personally confirmed that the NOAA version (Oct 2007 edition) is updated to May 2007 for most USHCN stations. There was a USHCN update in May or June 2007 which updated to late 2006 for most USHCN stations.
I personally confirmed that the most recent GHCN version (ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/v2 Feb 2008 edition), as noted by Jerry, contains USHCN updates only to March 2006 or so. Thus, GHCN is two USHCN updates behind at present (May 2007, Oct 2007) and its USHCN version is at least 14 months stale relative to what it could be using.