Introduction:

Present:

NCEP/NCAR Reanalysis (1948-1998+)

first attempt at a reanalysis covering such a long period of time
more "advanced" than the then-current operational forecast model
data is used by many people for varied purposes
produced by people with a NWP background
a large project: human errors
not the final word

Future Trends:

Better models

improvements in parameterizations (cloud, radiation, precip, etc)
improvements in the subsurface model
radiation code run more frequently (3 hours in N/N reanalysis)
higher horizontal resolution (Indian monsoon, orography)
higher vertical resolution (stratosphere)

Better assimilation techniques

better theoretical techniques (ex 4D variational method)
assimilation of rainfall (development in ETA)
direct assimilation of radiances (in MRF)
(better) use of SSM/I data
"soil analysis"
Since much of the satellite data was not used in the N/N Reanalysis,
there is a potential for a major improvements in the analyses.