Observations are central to the production of any reanalysis. Although there are a plethora of atmospheric and oceanic observations (particularly since the late-1970s), the observing systems themselves have always been deployed for other purposes, such as aviation, hydrology, and forecasting. In other words, reanalyses basically exploit observations that were collected for entirely different applications. Ironically, this is one of the strengths of the reanalysis method: it can create a coherent picture of the climate system from a disparate collection of historical data from a myriad of observational platforms. Although the input data vary from one reanalysis to the next, the observations themselves generally fall into one of two categories: conventional ("in situ") data and satellite records.
Conventional data are typically collected "on site," meaning that the sensor is present in the environment being observed. Conventional data for the surface originate from observing stations on land and from ships, buoys, and floats over the ocean. Above-surface data originate from radiosondes, pilot balloons, dropsondes, and aircraft. Wind profiler observations and manually derived pressure estimates also loosely fall into the conventional data category.
Satellite observations themselves are often grouped into two broad categories. The first category consists of observations from the TIROS Operational Vertical Sounder (TOVS) and its replacement, the Advanced TOVS, a series of polar-orbiting satellites that contained three separate instruments: the High Resolution Infrared Radiation Sounder (HIRS), the Microwave Sounding Unit (MSU), and the Stratospheric Sounding Unit (SSU). The second category consists of observations from a host of other satellite systems that measured one or more environmental elements, such as wind (e.g., GOES, GMS, METEOSAT, MODIS, SSM/I) or rain rates (e.g., SSM/I, TRMM). Other commonly used satellite data include AQUA-AIRS, METOP-IASI, CHAMP/COSMIC, ERS, and SBUV.