The NCEP reanalysis uses a global data assimilation system which is kept unchanged over the period 1957 through 1999 (Kalnay et. al, 1996). A T62 spectral resolution with 28 sigma levels, seven of them below 850 hPa, is used. Convective parameterization is performed by a simplified Arakawa-Schubert convective scheme (Pan and Wu, 1994). The NCEP reanalyses are not initialized because the statistical spectral interpolation eliminates imbalances in the initial state. The data base has been enhanced with ship, rawinsonde, aircraft and satellite data.

The study of Stendel and Arpe (1997) reveals that the consistency of both NCEP and ERA reanalyses is better than in the operational analyses. However, in the tropics and extra-tropical summer, NCEP and ERA reanalyses overestimate the convective precipitation.

The snow problem: snow matrix for 1973 had been kept for the interval 1974-1994.The problem was fixed in another run, from 1979 (NCEP/DOE reanalysis).