The NCEP/NCAR (National Center for Environmental Protection/National Center for Atmospheric Research) Re-analysis 1 project (NCEP R-1) is an analysis/forecast system to perform data assimilation using past data from 1948 to the present. In 1998 NCEP/DOE (National Center for Environmental Protection/Department of Energy) partnered to produce the Re-analysis 2 project (NCEP R-2) to address model and human errors identified in the old reanalysis R-1 model. Several error fixes were introduced in going from R-1 to R-2, most importantly the Southern Hemisphere bogus data problem (1979-1992) that rendered R-1 almost unusable for Southern Hemisphere extratropical investigations for the affected years. The oceanic albedo was too high by a factor of about two for the entire reanalysis period, may have affected the Southern more than the Northern Hemisphere because of the greater proportion of ocean surface. Other errors were found with the snow-cover analysis, discontinuities in the relative humidity and cloudiness relationship table and a snow-melt term and the presence of an undesired “spectral snow” feature. Output of the NCEP R-1 product continues to be released to the present day, however, R-2 has only been performed for the years of 1979-2002 for detailed evaluation. These datasets can be found on cherax in the subdirectories for R1 (other gauss, pressure, spectral, surface, surface gauss and tropopause), which match the names used by NCEP in their anonymous server. For R-2 other gauss and surface gauss have been merged into gaussian grid, and tropopause data do not exist at this point in time. All variables that are a function of time are resolved every 6 hours (0h, 6h, 12h and 18h UTC), and a complete year of data is written to a single file (either 1460 or 1464 times depending on whether the particular year is a leap year). The pressure data are on a 2.5 by 2.5 degree horizontal grid ( 144 × 73 regular grid points) for each of the 17 pressure levels2. The surface and tropopause data are also on a 2.5 by 2.5 degree horizontal grid (144 × 73 regular grid points). The surface gauss and other gauss (or gaussian grid in R-2) data have an approximate 1.8 by 1.8 degree horizontal grid (192 × 94 Gaussian grid points or approximately 200 km horizontal spacing). The spectral data is in coefficient form, there are 4032 for the T62 model resolution. Seasonal and monthly climatologies are available for the period 1948-2003 for the R-1 data and 1979-2001 for the R-2 data. For every variable there are four types of average netCDF files that are referred to as long-term monthly (lm), longterm seasonal (ls), year-by-year monthly (ym) and year-by-year seasonal (ys). The four seasons are referred to with the abbreviations DJF (December, January and February average), MAM (March, April and May average), JJA (June, July and August average) and SON (September, October and November average).