These are the "traditional" analysis products. The fields include: horizontal winds, omega (dP/dt), geopotential height, specific/relative humidity, absolute vorticity and divergence, The data are on standard pressure levels every 6 hours.
Data on isentropic surfaces include: horizontal winds, mass-weighted horizontal winds, omega (dP/dt), temperature, potential vorticity, relative humidity, Montgomery stream function, Brunt-Vaisala freq. squared. Not forgotten is the potential temperature at the surface. This data is available on 10 isentropic surfaces every 6 hours.
Gridded data on sigma surfaces include: rel. vorticity, divergence, temperature, specific humidity, horizontal winds, surface pressure, geopotential height. This data is available on the 28 model sigma levels every 6 hours.
The atmospheric analysis is available as a spectral sigma file. This is the only file with full precision data. For most problems, the truncation used by the GRIB files are reasonable. (For example, the temperatures are stored to the nearest 1/10 of a degree which is better than your average thermometer reading.) However, for perfectionists, people computing budgets, and those needing higher vertical resolution should go to the original spectral sigma files. Now the bad news. The NCEP post-processor runs on the cray. While a version will run on the SGI, it uses the cray to convert the file from IBM to ieee. Before you give up hope, people have been reading and using the spectral sigma data without using NCEP's post-processor. Not difficult but not something a new user will knock off in a hour.
Making the spectral data more accessible to the outside world is something we would like to do. However, it is low priority and will remain a low priority until there is outside demand. AFAIK there has been (* 1 *) request for easier access to the data.
One huge database is the diabatic heating and sub-grid momentum fluxes on the 28 sigma levels every 6 hours. One reason this set of data is so huge is that the diabatic heating is divided into its various components such as long wave heating. Interesting stuff but don't try it at home (unless you have n gigabytes of disk space and a few tape drives).
Quantities include: LW/SW cloud forcing, clear/cloudy, LW/SW, upward/downward fluxes.
Fields include: total cloud cover, convective and total precipitation.
The other fields include: surface wind stress, latent/sensible heat flux, soil temperature/moisture, gravity wave drag, SST, 2m temperature, 2m humidity, 10m winds, runoff, mean sea level pressure, surface pressure, and snow. The above list is not complete.
Other Reanalysis products include the first guess, hindcast forecasts (the skill of the forecasts give some idea about the quality of the analyses), BUFR data (the raw data from aircraft, rawinsondes, satellite, etc), optimal averages, statistics of the raw data utilization, and even a GCM simulation (sea-ice, SST boundary conditions). If you don't trust those NWPers with their fancy models and complicated equations on how to data assimulate, you can get the 'real' data (rawinsonde) from the BUFR files.
The above list are not complete. Quantities of limited interest such as x-gradient of log surface pressure, relative humidity between 0.44-0.72 sigma levels, "Best" 4-layer lifted index have not been mentioned.
Most of the above data were written in GRIB which can be quite easy to read.
Comments: Wesley.Ebisuzaki@noaa.gov, Wei.Shi@noaa.gov