I calculated the EOFs of monthly anomalies, all calendar months for 1948-2002, and although the leading EOF is the annular mode, the second and third modes are less recognizable. The leading EOFS explain 29, 12, and 10 percent of the variance. big GIF | PNG | PS

The documentation at the NCAR WWW page for this data set describes latitude circles erroneously shifted in a few months and a few months of isolated bad values (bulls-eyes). I examined these months (February 1902 to July 1909, inclusive, and December 1974), and fixed these problems. This data set I call "slptrenberth18992009.nc" and I am responsible for all errors in this netCDF file.

The earlier parts of the record had no values for the North Pole nor for the 75N and 85N latitude circles. I took "slptrenberth18992009.nc", interpolated values for these grid points, and wrote a netCDF file of the corrected, interpolated data, "slptrenberth18992009.interp.ncc". I am responsible for all errors in this netCDF file.

Even after these corrections there are still gridboxes and times with no data.

big GIF | JPG | PDF | PNG | PS

The reference for this data set is Trenberth and Paolino (1980), and the user is also encouraged to look at the documentation on the NCAR WWW page for this data set. The data set is derived from station data that has been processed by a variety of different methods, and the data from 1994 to the present is from the NCEP final analysis. December 1944 is filled with missing value flags.

Data (netCDF files)
Read the above text to see what the files are.