Thermometers thru the ages

After Climate gate broke, I got interested in seeing what a real plot of historical temperatures might look like. I found the NOAA data files at ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/v2/. Net rumor from the various global warming sites says the NOAA records are the primary sources, the temperature records at other institutions, like the recently hacked Hadley CRU, are said to be derived from the NOAA files.
This is an FTP site, Firefox will take you there but it can only download files one at a time. Filezilla, a free ftp program, can be told to download them all in one fell swoop.
NOAA stores the files zipped, and after FTPing them onto my computer and unzipping, I found the main file, v2.mean, is 44 megabytes of "mean" temperature records. The file is oldfashioned, showing its IBM punch card ancestry. It's what you get for reading boxes and boxes of punch cards onto a 9 track magnetic tape, and later posting the tape on the web. It's in ASCII, and each record is exactly 76 characters long, a 12 character station number, four character year, and 12 five character fields for each month's mean temperature. No spaces, comma's or separators between the fields, which is old fashioned punch card style.
I wrote a short program in C to sort the temperature readings out by year and compute the average temperature for each year. I compute the plain average (the one we learned in grade school) of all temperatures for each year. The earliest year is 1701, the last 2009. In the early years there are only 30 or 40 separate readings, by 1989 the number of readings swells to 100,000 per year, and then declines to about 14,000 in 2009. Either the NOAA funding for data entry ran short, or someone stopped recording (or removed) temperatures from stations he deemed unimportant or inaccurate. Some net commentators have claimed that stations in cold locations and higher altitudes were deliberately removed to make the warming trend more obvious.
I plotted the results in Excel, and sure enough, the rather fuzzy plot shows world average temperature at 9 or 10 degrees Centigrade in 1701, rising to maybe 14 degrees Centigrade in 2009. 1701 is deep in the little ice age. Plus, in the early years, the 30 to 40 reading come from temperate-to-chilly locations like Britain, Germany, France and the American colonies. By the peak in 1980 there are a lot or readings from steamy hot places like the Sahara desert and the Amazon basin.
We see maybe 4 to 5 degrees warming over 308 years, call it 1.5 degrees Centigrade per century.
There is a lot more to do. There is a max temp and a min temp file to be plotted, and then there are files of "adjusted" data. It will be interesting to see what they look like. I ought to convert the temperatures to Fahrenheit which has more juice for Americans. I ought to compute the standard deviation of the temperature readings, and add some error checking code to spot way out of whack ridiculous temperatures.