A 3–D animated image of Reanalysis-2 data for the first 10 days of July 1979, in six-hourly intervals. This animation shows a constant 100 mph wind speed surface in red. (Note the stronger, more widespread, polar jet stream in the Southern Hemisphere—this is July, during the southern winter.) A cyan-colored, constant temperature sheet of 0°C ripples across the globe, showing the freezing level. Near-surface wind flow is denoted by white flowlines. This image was generated with plots from Unidata's Integrated Data Viewer combined with ImageMagick.
Reanalysis of past weather data presents a clear picture of past weather, independent of the many varieties of instruments used to take measurements over the years. Through a variety of methods, observations from various instruments are added together onto a regularly spaced grid of data. Placing all instrument observations onto a regularly spaced grid makes comparing the actual observations with other gridded datasets easier. In addition to putting observations onto a grid, reanalysis also holds the gridding model constant—it doesn't change the programming—keeping the historical record uninfluenced by artificial factors. Reanalysis helps ensure a level playing field for all instruments throughout the historical record.
The following reanalysis datasets are available through NOAA's National Operational Model Archive and Distribution System (NOMADS).
Climate Forecast System Version 2 (CFSv2)
CFSv2 provides a global reanalysis (a best estimate of the observed state of the atmosphere), a global reforecast of past weather down to 0.5 degree horizontal resolution, and an operational, seasonal forecast of weather out to nine months.
North American Regional Reanalysis (NARR)
NARR is a regional reanalysis of North America containing temperatures, winds, moisture, soil data, and dozens of other parameters at 32 km horizontal resolution.
Reanalysis-1 / Reanalysis-2
Reanalysis-1 / Reanalysis-2 are two global reanalyses of atmospheric data spanning 1948/1979 to the present at a 2.5° horizontal resolution.