GISS gridded data is online but in a format that is unintelligible to people who are working with modern computer languages, as opposed to Fortran and who do not know whether their machines are “littleendian” or “bigendian” (see here for GISS discussion) – phrases rather reminiscent of Gulliver’s Travels, perhaps an apt text for Hansen. (For readers unfamiliar with Gulliver’s Travels, at one point in his travels, Gulliver was called upon to intervene in a war between Lilliputians over whether boiled eggs should be cut first at the big end or the little end.)
Thanks to the resourceful Nicholas, the GISS binary files can now be accessed in R and thus to any other modern languages. There are 3 gridded data files in the GISS directory tp://data.giss.nasa.gov/pub/gistemp/download, one pertaining to 250 km, one pertaining to 1200 km and one to HadR2 (which I haven’t identified yet.) The functions here succeed for all three files.
Unlike Jones who divides the world into 2592 five degree lat-long boxes, Hansen divides the world into 8000 equal area boxes (whose N,S,E and W limits have to be specified.)
The following script will enable a reader to download one of these GISS binary files and organize the information into an information file grid.info and a list of 8000 monthly time series giss.grid (mutatis mutandi for the other files.) This takes a few minutes as the files being downloaded are 20-40 MB.
handle < – file("SBBX.Tsurf250", "rb");
Data <- read_GISS_data_SBBX(handle);
To plot the gridded data for the 250-km smoothed box containing Wellington NZ, you can do the following
# Next LATS LATN LONW LONE NR1 NR2 DL
#6719 1548 -42.07 -39.79 174 177 4 1442 1116772928 -40.93 175.5
There are some mysteries in connection with the gridded data. At present, I can’t get from the station data to the gridded data, though I can get fairly close.
Not unexpectedly with code as crappy as this, it seems to require considerable ongoing maintenance. In the version of toSBBXgrid.f downloaded by Arthur Edelstein only a few weeks ago, there was a line as follows:
PARAMETER (IYREND=2007, IYBASE=1951,LYBASE=1980)
Obviously it makes no sense to specify IYREND at 2007. In the version presently online, this has been changed to parameterize the closing value – which is an improvement, but you’d think that this would have been done some time previously in the past 20 years.
I’m really stumped by the following problem: GISS gridded data has values for the year 1880, but the corresponding station data begins only in 1881. Where does the 1880 data come from? Or did I get wrongfooted somewhere? Suggestions welcomed.
Update: It looks like the “1880 problem” is only a problem arising from the web presentation which starts in 1881 even though the underlying data series start in 1880. Until a couple of weeks ago the only way of getting at GISS data was scraping the web pages (GISS refusing to make any sensible format available.) However, there are now some more improved versions available online and embedded in them are series starting in 1880.
Here’s what I’ve been able to emulate so far for one test case: