A type of consolidation of weather and climate information where past weather records are run through complex data assimilation models — originally designed
for weather forecasting — to produce a single, uniform global data set for
50 years or more.
According to Paul Edwards in his book A Vast Machine:
"Traditional climate data consist mostly of averages for single variables (temperature, precipitation, etc.) over periods of a month or more. Reanalysis produces a much different kind of data: all-variable, physically consistent data sets containing information for millions of gridpoints every six hours. Although biases in the models prevent them from displacing traditional climate data, climate statistics calculated from reanalysis data can reveal "fingerprints" of climate change not detectable in traditional data."